<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Hamza Mahmood on Medium]]></title>
        <description><![CDATA[Stories by Hamza Mahmood on Medium]]></description>
        <link>https://medium.com/@hamzamahmood?source=rss-9a046831034------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*wbIHe3vp8ufQJpAjhkigAw.jpeg</url>
            <title>Stories by Hamza Mahmood on Medium</title>
            <link>https://medium.com/@hamzamahmood?source=rss-9a046831034------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 06:13:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@hamzamahmood/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[That Pakistani meme which shook the world.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://hamzamahmood.medium.com/that-pakistani-meme-which-shook-the-world-cd09dc8f1217?source=rss-9a046831034------2"><img src="https://cdn-images-1.medium.com/max/1000/1*0Jmf9Ny6Nmw0lfB6GUux9Q.png" width="1000"></a></p><p class="medium-feed-snippet">A retrospect of how one humorous meme got sold for a hefty price, how NFTs work, what artists can expect in Web 3.0?</p><p class="medium-feed-link"><a href="https://hamzamahmood.medium.com/that-pakistani-meme-which-shook-the-world-cd09dc8f1217?source=rss-9a046831034------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://hamzamahmood.medium.com/that-pakistani-meme-which-shook-the-world-cd09dc8f1217?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd09dc8f1217</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[memes]]></category>
            <category><![CDATA[ethereum]]></category>
            <category><![CDATA[metaverse]]></category>
            <category><![CDATA[nft]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Tue, 31 Aug 2021 04:33:49 GMT</pubDate>
            <atom:updated>2021-08-31T04:35:04.617Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Pakistani “E-Commerce” User Experience]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://hamzamahmood.medium.com/the-pakistani-e-commerce-user-experience-c8dbd9d8a9ef?source=rss-9a046831034------2"><img src="https://cdn-images-1.medium.com/max/2064/1*fK59XE6mtlb6sdItiBPMag.png" width="2064"></a></p><p class="medium-feed-snippet">Why technology becomes a bottleneck for Daraz and Cheetay to achieve great UX? Exploring where the gap is, how product backlog and systems&#x2026;</p><p class="medium-feed-link"><a href="https://hamzamahmood.medium.com/the-pakistani-e-commerce-user-experience-c8dbd9d8a9ef?source=rss-9a046831034------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://hamzamahmood.medium.com/the-pakistani-e-commerce-user-experience-c8dbd9d8a9ef?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8dbd9d8a9ef</guid>
            <category><![CDATA[design]]></category>
            <category><![CDATA[systems-thinking]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[pakistan]]></category>
            <category><![CDATA[ux]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Sun, 06 Dec 2020 22:04:35 GMT</pubDate>
            <atom:updated>2020-12-06T22:04:35.689Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Uncovering design problems by asking the right questions]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://uxdesign.cc/uncovering-design-problems-by-asking-the-right-questions-4469f2ba4772?source=rss-9a046831034------2"><img src="https://cdn-images-1.medium.com/max/2600/1*bBzX39w9E6tLqVp1kyREzA.png" width="2732"></a></p><p class="medium-feed-snippet">The internet and the onset of the fourth industrial revolution have brought unique set of challenges in the space of human-centered design.</p><p class="medium-feed-link"><a href="https://uxdesign.cc/uncovering-design-problems-by-asking-the-right-questions-4469f2ba4772?source=rss-9a046831034------2">Continue reading on UX Collective »</a></p></div>]]></description>
            <link>https://uxdesign.cc/uncovering-design-problems-by-asking-the-right-questions-4469f2ba4772?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/4469f2ba4772</guid>
            <category><![CDATA[user-research]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[user-experience]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Mon, 16 Sep 2019 13:12:17 GMT</pubDate>
            <atom:updated>2019-09-16T21:42:00.544Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[I felt the Amazon app was outdated, so I redesigned it using Design Thinking — a UX case study]]></title>
            <link>https://uxdesign.cc/i-felt-the-amazon-app-was-outdated-so-i-redesigned-it-using-design-thinking-a-ux-case-study-549c5880d9be?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/549c5880d9be</guid>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[case-study]]></category>
            <category><![CDATA[ui]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[ux-design]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Fri, 14 Jun 2019 22:46:05 GMT</pubDate>
            <atom:updated>2019-06-24T23:38:41.490Z</atom:updated>
            <content:encoded><![CDATA[<h3>I felt the Amazon app was outdated, so I redesigned it using Design Thinking principles — a UX case study</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ohPlheUesEn3O0sgEiAtiA.png" /><figcaption>Workstation @ the Stony Brook lab</figcaption></figure><p>I have been an avid user of the Amazon app for as long as I remember, both its web and mobile version. It has conveniently assisted me in purchasing stuff which I otherwise would not find in stores, especially rare books, even sneakers that suit my style and more.</p><p>With time as I gained more experience as a UX engineer, I started closely observing the additions being introduced into the app with a multitude of features that most people never got to discover (Amazon Programs, the scan item feature among other). One thing, in particular, bothered me about the mobile version: the information clutter present in most of their pages and the element of newness in their UI which the app severely lacks. So, I picked my pencil and piece of paper then started jotting down ideas that could potentially improve the usability of the app.</p><p><strong>The Challenge: </strong>Pick<strong> </strong>a mobile application which I am familiar with and by utilizing measurable data from existing users and inspiration from modern visual layouts, try to optimize the app by making it accessible for users through a cleaner UI, reducing clutter and improving user flow.</p><p><strong>The solution: </strong>Considering the factors and the user feedback obtained through the usability study I conducted, I propose a modified version of the Amazon mobile app that follows Amazon’s timeless branding, structurally arranged information (product details, description, and reviews), prioritizing the powerful ‘scan item’ feature and easier side-to-side comparison of items. This study is meant to provide insight into what aspects within Amazon’s mobile app can be improved or corrected to address the pain points of users, given the changes can significantly impact the company’s revenue generation with the paradigm shift involving people moving towards digital modes of shopping.</p><h4>Amazon: A Bird’s Eye View of “The Everything Store “</h4><p>Amazon.com is a widely popular platform used by people to purchase products based on their needs and wants. The application boasts one of the largest catalogs of products that range from books, clothes, electronics, toys to household items and makeup. The company started out in 1994 and currently is valued at $888.8bn, having hit the trillion dollar mark earlier in September 2018.</p><p>According to the Business Insider report; in 2018 alone, users spent an average 12.8 billion minutes per month (84.5% of total usage) in the Amazon app compared to 2.8 billion minutes per month (15.5% of total usage) on the browser version of the app, showcasing the massive conversion rate achieved through the mobile app.</p><p>Talking about the mobile app, it has loads of features that users get to discover with time but they are so many that most of them end up not being even touched. The app is regularly updated with the addition of advanced features and potential bug fixes that go towards enhancing the overall user experience. Each category of products has its own set of options to choose from which adds to the app’s complexity so there is no <em>one-size-fits-all</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*4sfJx6mVG4BoZ8u0.gif" /><figcaption>Too many features add absurd levels of complexity to the app which Amazon effectively handles. <a href="https://i.pinimg.com/originals/eb/a2/f0/eba2f05d6c3f41f347f00377a6921e9c.jpg?__hstc=246535899.1289f84a362c41b80e5e8776d3502129.1435233910711.1436779825313.1436783333835.39&amp;__hssc=246535899.36.1436783333835&amp;__hsfp=4289683868">Image Source</a></figcaption></figure><p>However, despite the massive user base drawn by the company, there is room to further improve the core workflow of the app that will go a long way in securing and retaining users, especially those within the younger age bracket.</p><blockquote><em>As an aside, I did not have any access to the massive data that Amazon has collected over the years. All the decisions their development team has made is backed by years of analytics on user behavior, navigation, and purchase trends. With that in mind, the concept I present is not exhaustive and is solely based on data collected through guerilla usability testing.</em></blockquote><h4>Personal Goals</h4><p>I was determined to carry out the project with the following goals in mind:</p><ul><li>To create a more enjoyable, accessible and clutter-free shopping experience.</li><li>Drive UI design through <a href="https://www.interaction-design.org/literature/article/design-thinking-getting-started-with-empathy">empathy</a> and Design Thinking.</li><li>Create designs which are backed by data from the usability study. Use Proto.io to create refined, high fidelity mockups and</li><li>Keeping the prototype consistent with Amazon’s brand identity</li></ul><p>For the duration of the project I played the part of the following roles 1) User Researcher, 2) Data analyst, 3) UI/UX visualizer, 4) Product designer</p><h3>Course of Action</h3><p>As a starting base, I devised the project schedule, breaking down the project into milestones. I initiated a usability test to identify how key features are used within the current Amazon app framework and where there are shortcomings. I also tried to understand what features users are not aware of, in order to prioritize the feature content on the app for more visibility. The process below entails how I approached the problem statement and using the data gathered push towards updating the task flow and making amendments to the current UI for improved navigability and user experience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hXp1LjCuMv37uMwdYWk05w.png" /></figure><h3><strong>User Research &amp; Usability Study</strong></h3><p>In order to understand a product’s user is to go meet them, talk them through the <strong>goals</strong> they want to achieve with the app and to empathize with them with regards to the <strong>pain points </strong>they share. I had the privilege to use Stony Brook’s computer labs to stage interviews for the usability study, bringing in 5 participants who were willing to share their experience with the app.</p><p>The purpose of the usability study was to identify bottlenecks in the current version of the app and funnels where the user is most likely to drop out of the app. So, I designed a short focused survey to gather preliminary data and feedback about:</p><ol><li>What are the demographics of the user? — age, gender, personality</li><li>What category of products they usually shop for?</li><li>Whether they research on other online/offline platforms before making a purchase?</li><li>How would they rate their experience on the web and mobile version of the app?</li></ol><p>After this, I carried out a timed test in which users had to perform set tasks which also involved using newer features they may not be aware of. The total time per interview took 30 mins.</p><h4>Analysis &amp; Findings</h4><p>The users were open to providing their feedback about the visual design of the app. The general verdict declared the current version as outdated and the content placement, especially within the home and product pages, to be cluttered and sometimes overwhelming. The survey also supported that the average number of users are<strong> in favor of a UI makeover</strong>.</p><p>From the 5 conducted interviews and 30 responses from the survey, I derived the following analysis:</p><ul><li>40% of the responders fell between the 18–24 age bracket while 53.3% fell in the 25–34 bracket.</li><li>shop an average 2–3 times per month on Amazon (platform agnostic),</li><li>are accustomed to using modern technology and find it easier to adopt newer technology.</li><li>22 of 30 responders consider themselves as introverts.</li><li>26 of 30 users would first research and read through the reviews before making a confirmed purchase</li></ul><p>Some interesting findings from the survey:</p><ul><li>Household items are the most commonly shopped with 33.3% while the others buy electronics (30%)</li><li>80% of the responders said that they never got around to using the ‘camera scan’ feature or did not feel the need to.</li><li>60% of the responders associated <em>convenience</em> as the prime reason for using Amazon, 16% said <em>navigability</em> while 14% went for a <em>variety of options</em>.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hPLSXTUQNk_Dx_832_T-Wg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qJYWQkL_6Np0ULuNYi0xGA.jpeg" /><figcaption>Goals and Pain points established from user responses</figcaption></figure><p><strong>User Goals:</strong> The goals Amazon users want to accomplish mainly revolved around the experience of buying a product as seamlessly as possible, potentially getting a bang for their buck. They want to save time going to a physical store and spending long hours traipsing through to find what they need. Also, when browsing through a product, a typical user would ideally check for reviews and see what others say before making a decision.</p><p><strong>Pain points: </strong>Users shared a common sentiment that there is an overabundance of reviews on the product pages, irrelevant recommendations on their home feed and too many features that they do not particularly need.</p><h4><strong>User Persona &amp; Branding</strong></h4><p>The diagram below is a depiction of what a typical Amazon user’s persona looks like based on the usability study and survey data gathered through the process:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qdQcrMGJJYMf4-7qxTTw4Q.jpeg" /><figcaption>User Persona board containing data from the usability study.</figcaption></figure><p>Talking about the branding, the icon redesign above embraces the sleeker look adopted by several companies such as Instagram, Skype and Apple Music. Amazon presents itself as a technology company that continuously strives to become pioneers of new technology whether it is in the shopping space or in cloud technology. Even now the company is inching closer towards space travel. So it is not hard to imagine the company to change the outlook of the app if necessary.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*dg2WQ9wfHL92W1infYiUhA.png" /><figcaption>The Amazon Shopping App Icon, Reimagined</figcaption></figure><p>The color palette below is resonant of Amazon and produces a mood of warmth, happiness, and joy which is associated with shopping at their online store. The darker colors are representative of the company’s corporate outlook, nicely balancing out the brighter colors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kDpFF5b20XF9fsQ-cU_IiA.png" /><figcaption>Amazon’s original brand colors are being utilized</figcaption></figure><h4><strong>Hi-fidelity Prototype</strong></h4><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.behance.net%2Fgallery%2F79601397%2FAmazon-Mobile-App-Redesign-Concept-2019%3Fiframe%3D1&amp;url=https%3A%2F%2Fwww.behance.net%2Fgallery%2F79601397%2FAmazon-Mobile-App-Redesign-Concept-2019&amp;image=https%3A%2F%2Fmir-s3-cdn-cf.behance.net%2Fprojects%2F404%2Fed4c7c79601397.Y3JvcCwxNDAwLDEwOTUsMCww.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=behance" width="700" height="900" frameborder="0" scrolling="no"><a href="https://medium.com/media/4a6e8ceae870c0fc80d7c2a9c1571e14/href">https://medium.com/media/4a6e8ceae870c0fc80d7c2a9c1571e14/href</a></iframe><h3>Part 1 of 4 — Reinventing the Product Page 📦</h3><p>In a content-heavy app where the structural organization of information is critical to helping a user’s decision in purchasing a product. The workflow (showcased in the figure below) is designed to cater to the pain points exhibited by users.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i_ZM84PX_auQRHP1wp0lBg.png" /></figure><ol><li>The taskbar includes updated icons on a white background to give a more modern look</li><li>The price range is set to bold to make it stand out for users. The <strong>See Shipping Options</strong> link on the right navigates to the list of products with the mentioned prices.</li><li>In order to choose a color, the user has to make extra clicks to choose the appropriate options before adding it to cart. In this modified version, the user can slide through the options sideways without navigating away from the screen or opening a popup to highlight the color and size of choice.</li><li>The same mechanism is applied in selecting the size of the shoe.</li><li>The clickable arrow button opens to a task panel of two buttons that allows the user to add to a <strong>wish list </strong>or to <strong>compare</strong> with similar looking products.</li><li>To address the challenge of information overload/clutter, the redesign makes use of accordion dropdowns that change state as the user clicks on it to display details.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a7s1FFySVQEc1zeps-luCw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RyhFCQ6AkXGI2ncKQOjs8g.png" /><figcaption>UI modification for the action buttons. Small changes can make big differences.</figcaption></figure><h3><strong>Part 2 of 4 — Optimizing The Checkout Process </strong>🛒</h3><p>With the proliferation of e-commerce platforms and digital marketplaces, companies are pushing for more seamless ways to optimize the checkout process. Amazon’s checkout process is at least effective if not the most aesthetically appealing to the eye. Also, the app includes recommendations within the checkout process which are usually ignored by users, falling into the ‘out of sight, out of mind’ conundrum.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6A10sqGwE7iUftticohJfg.png" /></figure><p>So in order to solve this issue, the <em>Checkout</em> page is kept clean by including only the most essential information without the need for the user to scroll down for recommendations. The user can update the quantity of the item, save it for later or remove it altogether. The order can be specified if it contains a gift item by checking the box. Upon clicking the checkout button, users have full visibility about which stage of the checkout process they are in, indicated by the circular navigation buttons.</p><p>In the <em>Shipping</em> section, the user can check the desired address where the items are to be delivered. There is also the ability to add a new address or selecting the nearest pickup location.</p><p>The <em>Payment</em> section is kept simple as well with the user selecting the payment option or add a new one. The user also has the convenience to scan their credit or debit card with a click of a button.</p><p>In the <em>Place Order </em>or<em> Review </em>section, only the most necessary details are displayed. Ample space is kept between the UI components for better readability.</p><h3>Part 3 of 4 — Prioritizing the Scan Item button 📷</h3><p>One of the more powerful features that the app recently introduced is the camera feature, where a user scans an item or uploads an image of an item to get results that are the closest match also suggest other relevant items available at reasonable prices.</p><p>Given the usefulness and effectiveness of the feature, I have given the camera scan its own button and for increased discoverability placing it below the sliding services section. The feature has a massive potential to boost online sales as users can compare products when shopping in brick and mortar stores to get better deals on the Amazon app!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SEoodK9OwctR17vsM1AoEw.png" /></figure><p>As a plus, it is better that users are given a short tutorial on how to use the feature. Clicking the scan button will trigger a pop-up explaining how to use the feature, facilitating users to give it a test-drive at least once.</p><h3>Part 4 of 4 — Introducing Dark Mode 🌙</h3><p>During the usability study, I made a hypothesis that users falling into the younger age bracket are more likely to browse the app in dimly lit rooms. Apps such as Medium, Twitter are content-heavy and people read through readable material before going to sleep.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lhfAJ5vbsBJvMYG2y9tXJA.png" /></figure><p>The purpose of<strong> dark mode</strong> is two-fold in Amazon’s case:</p><ul><li>To save battery by utilizing the OLED screens of modern mobile phones</li><li>In order to make it easier on the eyes for users when they are browsing through products or reading product reviews, questions and answers.</li></ul><p>Participants were accepting of the idea that a night mode might actually increase their engagement with the app but there are trade-offs with respect to keeping a consistency in the app which can be a huge pain especially when dealing with images with lighter backgrounds.</p><h3>Reflection and Future Work</h3><p>This project presented a golden opportunity for me to work on my UX chops, getting into my creative zone to work on an interesting problem. I intend wholeheartedly to become a UX Engineer/product designer and during the past 4 months, I have learned to channel my creativity efficiently by understanding user requirements through empathy, working in an <em>agile</em> setting to iterate quickly in prototype creation, understanding bottlenecks within apps and eventually pushing myself to design a more optimal version of an app.</p><blockquote>Its not about building the thing right, its about building the right thing</blockquote><p>For future work, I would continue to gather more data about users and if possible onboard more candidates to test the current app in order to reduce existing errors and any confirmation bias.</p><p><em>I am working on more product design projects and creating magic. If you liked this read I have written another article on Typography:</em></p><p><a href="https://medium.com/startup-grind/how-typography-impacts-your-business-outlook-a-primer-3f092261d294">How Typography Impacts Your Business Outlook — A Primer</a></p><p><em>If you want to discuss UX and design related projects. Hit me up on </em><a href="https://twitter.com/mahmooyo"><em>Twitter</em></a><em> </em>😃</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=549c5880d9be" width="1" height="1" alt=""><hr><p><a href="https://uxdesign.cc/i-felt-the-amazon-app-was-outdated-so-i-redesigned-it-using-design-thinking-a-ux-case-study-549c5880d9be">I felt the Amazon app was outdated, so I redesigned it using Design Thinking — a UX case study</a> was originally published in <a href="https://uxdesign.cc/">UX Collective</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GlobalHacks VII, Conquered]]></title>
            <link>https://hamzamahmood.medium.com/globalhacks-vii-conquered-6ac57bb2739c?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ac57bb2739c</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[travel]]></category>
            <category><![CDATA[motivation]]></category>
            <category><![CDATA[hackathons]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Fri, 08 Mar 2019 19:17:57 GMT</pubDate>
            <atom:updated>2019-05-28T07:57:11.481Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vehwk1IAhyU7zBaQJRBi7A.jpeg" /></figure><h4>How a team of three brown dudes won $15,000 of cash prize at a Hackathon Event in St. Louis</h4><p>It was the break of autumn in my campus dormitory, reading Brad Stone’s highly descriptive book on Amazon, “The Everything Store”. The day started off well and was only going to get better as I got a call from my good friend Jay, who excitedly asked whether I had any commitments for the day.</p><p>I casually replied, “No, but what’s up? You seem excited!”</p><p>It was that time of the semester when my senior colleagues were actively looking for jobs and internships. So, I sheepishly thought he was about to let me in on getting an offer or something. However, the question he asked was unique.</p><p>“Do you want to attend a hackathon in St. Louis?”, he fervently asked.</p><p>Now, Jay had my attention. “Tell me more about it…”, I said.</p><p>While he explained the situation, the first thing that came to mind was that it had been ages since I last attended a hackathon, and secondly, there was a surge of excitement that flowed through my nerves as this was a chance to do something meaningful and productive.</p><p>“The winning prize is 15 grand, by the way!”, Jay added. This was it for me. This was too good an opportunity not to miss. Jay was already in touch with our soon-to-be third teammate, Abhishek. The plan was set in motion, we booked our flights, got done with our homework well in advance and were ready to depart to the home of the Cardinals.</p><h4>ENTERING ST. LOUIS</h4><p>The place where the team had to rendezvous was located in mid-west. To give a very brief summary of the place, the city was ranked as the 3rd worst crime rate in the country in 2017. Missouri , where the city is situated, is also considered as a ‘red’ state or conservative-leaning, with residents more susceptible to racism and bigotry. The folk back at home were not really in favor of me traveling there and for good reason, but that did not stop me.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f0ynfoJvRjRF6Pi7oqCrJw.jpeg" /><figcaption>Morning time at the arena. It got crowded to full capacity in the next 3 hours.</figcaption></figure><p>The event was set in a very nice part of town as a matter-of-fact. We were to spend 3 days in the Chaifetz Arena, which was the property of the fabled St. Louis University. The arena overlooked the gateway arch (see the picture below). The weather was slightly chillier than in New York and remained overcast for the duration of my stay there.</p><p>A moment of self-reflection here.</p><p>There were moments in time where I pictured myself connecting with people from all four corners of the United States, to meet the smartest engineers, designers, data wranglers of my generation. To get acquainted with the much talked about startup culture, inspired by Silicon Valley. This was the chance, here I was about to get what I dreamed of.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*e8Tn9uq1fTnXCf36wpI1Ww.jpeg" /><figcaption>View of the Gateway Arch from outside the Arena.</figcaption></figure><h4>THE PROBLEM STATEMENT</h4><p>As I fiddled with my Rubik cube to pass the time, the event head took the stage, shortly. The voice was coming from far off where we were sitting. It was an enthusiastic bellowing voice to which the participants responded with a massive cacophony of sounds.</p><p>After going through the rudimentary details, we got to the meaty part of the speech. The scope was to design an app or a system that assisted<strong> new immigrants to reach out to a community in times of needs,</strong> To promote a sense of <strong>social belonging. </strong>A platform that could <strong>foster local communities</strong> and help them <strong>go beyond geographic, lingual, economic and cultural boundaries,</strong> with the ability for people to<strong> share lived experiences.</strong></p><p>Once the briefing was done, what followed was a thrilling, fast-paced series of events which we were able to make it out of alive, with special thanks to sugarfree Redbull and finely brewed coffee.</p><h4>OUR APPROACH</h4><p>From the get-go, we wanted to be super effective with our time and started to discuss our roles. I was to handle the front-end and documentation part of the application, while Jay took care of integrating AI and writing service-level REST code and Abhishek chose to code the backend portion. No task was easier than the other as the presentation layer of the product was critical and how seamlessly it worked with the variety of users which it was to target.</p><p>Although we had 48 hours to build a solution, we first took out 4–5 hours to brainstorm and conceptualize the idea itself, asking mentors and engineers part of the industry for advice and key elements to consider when building the solution. We were highly tempted to start implementing the prototype but we chose not to and this was Jay’s call, and we as a team placed our trust in him.</p><p>First, we mapped out the workflow of the app, to explore what was already out there, where the user pain points existed and how to accomplish the user’s goal with as little intrusion as possible. Our understanding of the problem was that each ‘community’, represented by their faith, culture, and race had a particular messaging app of choice. For instance, people from the Chinese community are regular users of WeChat and they would not unnecessarily jump onto a different service. Also, immigrants who belong to an older age bracket have a limited understanding of such apps, so they will resort to using more traditional methods of communication such as SMS or phone call.</p><h4>WHAT WE DESIGNED</h4><p>It was not long before we found our Eureka moment. We were not to make any new app and instead decided to go ‘platform-less’. We innovated a system that integrated with existing technologies to help immigrants connect across multiple platforms. We implemented a ‘<em>one click</em>’ sharing of data, news, and events based on the preference of the user, who may choose to get relevant information from any part of the world or from their nearby current location in the States. This functioned like a subscription or ‘opt-in’ model where getting information is fast and easy.</p><p>The cross-platform nature of the system allowed communities to inter-mingle with other communities, to plan out social gatherings or to help find jobs. People in difficult position can take advantage of the system’s Multi-Lingual helpline that can be accessed from any popular communication app. The user can ask questions, talk about worries, seek advice/guidance, get doubts cleared by simply dropping an anonymous audio or text message to the helpline.</p><p>In the first pass, our AI system will try to sift through a repository of FAQs. If it is able to find the closely matched question, the autogenerated voice will send the user a possible solution to the query. If that does not work (considering the limitation of AI), the question will be forwarded to a verified person who is allowed to answer the query. At no stage is the person aware of who he/she is helping, to maintain anonymity.</p><h4>THE VERDICT</h4><p>There were three rounds in total. First one was the most nerve-racking as a mere 30% of the participants were going to pass through. As a group of judges came to our table, we were able to successfully explain our idea. We gauged their reaction and it seemed pretty much in the positive, receiving enthusiastic nods and surprised murmurs as we presented our idea.</p><p>We passed the first round with flying colors. That boosted our morale and the next round was the investors round, in which we had 10 mins to explain and showcase a demo of our prototype. We had a buffer time of 2 hours to test for any bugs and fix them before pitching to the judges. I was personally tense but I tried not to get the best of me. Like in the previous round, we got the judges turning their heads to each other over our idea.</p><p>If we passed this round, we were sure shot getting a prize at the very least. We did not expect to reach this far in the competition</p><p>Yet when the time came for the announcements, we made it to top 3. During this point, we were confident about winning it big. The last round involved the final 3 teams to pitch their idea to the audience and the positions were to be decided based on that. We gave it our best and the rest we placed our trust in God.</p><p>“The first position goes to: Team ConnectUS!” — the voice of the event-head reached the depths of my soul. We had officially won $15,000 in cash prize.</p><p>It was no small feat and we were proud of this achievement. My folks back home were ecstatic with joy and soon a barrage of congratulatory messages came my way as I broadcasted this win on social media. It was a satisfying feeling all in all.</p><h4>TECHNOLOGY STACK</h4><p>The system was composed of a chatbot built on all platforms, connected to a single backend powered by <strong>Twilio</strong>, a communication Platform-as-a-Service. We made use of an autogenerated voice and messaging system to assist new immigrants arriving in the US. We designed the system to support a bare minimum level of communication using SMS, MMS, or traditional phone call services. We used Twilio&#39;s free tier to build the chatbot and extend them across popular messaging applications such as Whatsapp, Facebook Messenger, and even WeChat.</p><p>To host our backend we used <strong>Google Cloud Platform, </strong>utilizing a Python-based environment as we wrote the application in <strong>Flask</strong>. The same technology was used to power our Natural Language processing and multi-lingual query processing. Our database, user management system is based on <strong>Firebase</strong>.</p><h4>CHALLENGES</h4><p>In my opinion, time was a big constraint since we had 48 hours to build a functioning application to woo the judges. Generally, the lifecycle of a well-curated app spans from weeks to months, or even years depending upon the complexity and dimensions of the problem being addressed. We had to bootstrap quickly, with little room for error in decision making and quickly code up our prototype.</p><p>We also faced technological challenges. Our ambition was to cover as many platforms as possible to showcase the potential impact of our system which required insane amounts of computational resources. Also, properly handling the AI engine was difficult given the lack of data available online. Our strategy also entailed handling different dialects of a certain language but as of yet, even the best NLP frameworks (GCP/AWS) struggle to give accurate results.</p><h4>KEY TAKEAWAY</h4><p>I feel that the whole event was a big morale booster. It made me realize the importance of collaboration, putting trust on your teammates to work towards a great looking solution.</p><p>The Hackathon allowed me to tap into my inner computer geek, to write code fast and accurately. But it also strengthened my belief in the importance of brainstorming and ideation; all design thinking techniques, that allow a team to create the ‘right’ thing in the first try. It was big leverage as we did not have to iterate through our prototype multiple times.</p><p>Finally, and most importantly, this episode brought with it a self-belief, that one could do much more than they imagine if they put their thoughts into actions. One needs to have a self-starter mentality to accomplish goals, however big or small they are. Just put yourself out there and do the best you can.</p><h4>LINKS</h4><p><a href="https://github.com/hamza121star/connectUS">https://github.com/hamza121star/connectUS</a><br><a href="https://devpost.com/software/some-name-for-global-hacks-7">https://devpost.com/software/some-name-for-global-hacks-7</a><br><a href="https://www.slideshare.net/HamzaMahmood9/connect-us">https://www.slideshare.net/HamzaMahmood9/connect-us</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ac57bb2739c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Gradient Descent]]></title>
            <link>https://medium.com/data-science/gradient-descent-3a7db7520711?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/3a7db7520711</guid>
            <category><![CDATA[neural-networks]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Wed, 02 Jan 2019 21:18:15 GMT</pubDate>
            <atom:updated>2019-01-03T13:45:43.395Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qGHMj4APQhNtBMB4YR2ptg.png" /></figure><h4>It is a slippery slope, but promise it gets better at the bottom4</h4><blockquote><strong>tl;dr </strong>Gradient Descent is an optimization technique that is used to improve deep learning and neural network-based models by minimizing the cost function.</blockquote><p>In our previous post, we talked about activation functions (link <a href="https://towardsdatascience.com/activation-functions-in-neural-networks-83ff7f46a6bd">here</a>) and where it is used in machine learning models. However, we also heavily used the term ‘Gradient Descent’ which is a key element in deep learning models, which are going to talk about in this post.</p><h4>Definition and Nomenclature</h4><p>Gradient Descent is a process that occurs in the <strong>backpropagation</strong> phase where the goal is to continuously resample the gradient of the model’s parameter in the opposite direction based on the weight <em>w</em>, updating consistently until we reach the <strong>global minimum </strong>of function <em>J(w)</em>.</p><p>To put it simply, we use gradient descent to minimize the cost function, <em>J(w)</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bP7EaVyPvBSRIxqgA9TbGQ.png" /><figcaption><strong>Fig 1: </strong>How Gradient Descent works for one parameter, <em>w</em></figcaption></figure><p>An analogy could be drawn in the form of a steep mountain whose base touches the sea. We assume a person’s goal is to reach down to sea level. Ideally, the person would have to take one step at a time to reach the goal. Each step has a gradient in the negative direction (Note: the value can be of different magnitude). The person continues hiking down till he reaches the bottom or to a threshold point, where there is no room to go further down.</p><h4>Mathematics</h4><p>Let’s formalize the analogy into an algorithmic form. We compute the activation for the incoming parameters, we carry out feedforward by taking the weighted sum of the activation and its bias. We extract the error term of the output sample by subtracting it with the actual ‘target’ value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ETudkFMzVEMsUrVD.png" /></figure><p>The gradient descent process is exhibited in the form of the backpropagation step where we compute the error vectors δ backward, starting from the final layer. Depending upon the activation function, we identify how much change is required by much change is required by taking the partial derivative of the function with respect to <strong>w. </strong>The change value gets multiplied by the learning rate. As part of the output, we subtract this value from the previous output to get the updated value. We continue this till we reach convergence.</p><p>In the code below, I wanted to highlight how one can write a simple code to get a visualization of how gradient descent works. Running this piece of code; using the Tanh activation function, we will observe the current value of <strong>10</strong> go down to a value of <strong>8.407e-06</strong> on the 10000th iteration, which is our global minima.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/0c8c0d57b2e8d625f4f5b2f867b882b8/href">https://medium.com/media/0c8c0d57b2e8d625f4f5b2f867b882b8/href</a></iframe><p>There are a number of gradient descent algorithms out there. I’ll mention a few below:</p><ul><li>Batch Gradient Descent</li><li>Stochastic Gradient Descent</li><li>Mini-batch Gradient Descent</li></ul><p>If you want to go into the technicalities of the more recent ones, I highly recommend going through Sebastian Ruder’s <a href="http://ruder.io/optimizing-gradient-descent/">article</a> on the topic.</p><h4>Exploding &amp; Vanishing Gradients</h4><p>In deep networks or recurrent neural networks, there are two known issues explained in a paper by Pascanu et al (1994) — of exploding and vanishing gradients. This happens when are doing back propagation as we iterate through the code, there is a chance that the normal of the weight matrices going beyond 1. If this happens, the gradient explodes but if the normal is below 1, the gradient vanishes.</p><p>If we want to visualize exploding gradients, you will encounter at least one of the problems:</p><ul><li>The model will output ‘Nan’ values</li><li>The model will display very large changes upon each step</li><li>The error gradient values are consistently above 1.0 for each node in the training layer.</li></ul><h4>Solution: Gradient Clipping</h4><p>The solution to the exploding and vanishing gradient problem, we introduce gradient clipping, where we ‘clip’ the gradients if it goes over a certain threshold represented by a maximum absolute value. Hence, we keep the neural network stable as the weight values never reach the point that it returns a ‘Nan’. In a coded implementation removing the clipped gradients leads to ‘Nan’ values or infinite in the losses and fails to run further.</p><p>The code below showcases how to perform gradient clipping. Given that we have a vector of losses and a learning rate, we are able to compute a vector of gradients which are next clipped based on a maximum L2-norm value, which in this case I have written as 5.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/25d89e30cd1241912ade07867f265b47/href">https://medium.com/media/25d89e30cd1241912ade07867f265b47/href</a></iframe><p>So at the end of the day, when a question is posed to data scientist of what optimizer to use in order to minimize loss, there are a couple of factors to consider:</p><ul><li>The size of the training dataset</li><li>How quickly do we need the training data to achieve convergence</li></ul><p>Link of the paper referenced: <a href="http://proceedings.mlr.press/v28/pascanu13.pdf">http://proceedings.mlr.press/v28/pascanu13.pdf</a></p><h4>Conclusion</h4><p>In this write-up, we covered a number of things: We covered what is Gradient Descent is and how it works in a neural network. We went through the mathematics involved and implemented a coded version of it. Lastly, we covered the issues involving gradient descent in the form of the vanishing and exploding gradient problem and discussed the solution for it using gradient clipping. In the next lecture, we will explore what activation functions are and how they are crucial in deep learning models, so stay tuned for that!</p><p><em>Spread and share knowledge. If this article piqued your interest, share it among your friends or professional circles. For more data science and technology related posts follow me </em><a href="https://medium.com/@hamzamahmood"><em>here</em></a><em>.<br>I am also available on </em><a href="https://www.linkedin.com/in/hmahmood93/"><em>Linkedin</em></a><em> and occasionally </em><a href="https://twitter.com/mahmooyo"><em>tweet</em></a><em> stuff as well. :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3a7db7520711" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/gradient-descent-3a7db7520711">Gradient Descent</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Activation Functions in Neural Networks]]></title>
            <link>https://medium.com/data-science/activation-functions-in-neural-networks-83ff7f46a6bd?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/83ff7f46a6bd</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Mon, 31 Dec 2018 02:10:43 GMT</pubDate>
            <atom:updated>2019-01-02T21:27:44.371Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kYgSAIvAgQdxKe66wAtM3A.png" /></figure><h4>The motive, use cases, advantages and limitations</h4><blockquote><strong><em>tl;dr </em></strong><em>The post discusses the various linear and non-linear activation functions used in deep learning and neural networks. We also take a look into how each function performs in different situations, the advantages and disadvantages of each then finally concluding with one last activation function that out-performs the ones discussed in the case of a natural language processing application.</em></blockquote><h4>Introduction</h4><p>Continuing on with the topics on neural networks. Today we are going to discuss what <strong>Activation Functions</strong> are and try to understand the underlying concept. We do a deep dive into both linear and non-linear versions of these functions and explore where they work best and in where they fall short.</p><p>Before we move on what activation functions are, lets first refresh our brains on how neural networks operate. So, a neural network is a complex mesh of artificial neurons that imitates how the brain works. They take in input parameters with its associated weights and biases then we compute the weighted sum of the ‘activated’ neurons. Our activation function gets to decide which neurons will push forward the values into the next layer. How it works, is what we are going to see next.</p><h4>The Underlying Concept</h4><p>In the forward propagation steps, our artificial neurons receive inputs from different parameters or features. Ideally, each input has its own value or weight and biases which can display interdependency that leads to change in the final prediction value. This is a phenomenon referred to as the <strong>Interaction effect</strong>.</p><p>A good example would be when trying a regression model on a dataset of diabetic patients. The goal is to predict if a person runs the risk of diabetes based on their body weight and height. Some bodyweights indicate a greater risk of diabetes if a person has a shorter height compared to a taller person who relatively has better health index. There are of course other parameters which we are not considering at the moment. We say there is an interaction effect between height and bodyweight.</p><p>The activation function takes into account the interaction effects in different parameters and does a transformation after which it gets to decide which neuron <em>passes forward</em> the value into the next layer.</p><h4>Linear function</h4><p>We start off with the simplest function; the <strong>linear function</strong>. The value of <em>f(z)</em> increases proportionally with the value of <em>z</em>. The input value is the weighted sum of the weights and biases of the neurons in a layer. The linear function solves the issue of a binary step function where it reports only a value of 0 and 1.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FazlBgIgX3F5esU9l2uk5g.png" /><figcaption><strong>Fig 1: </strong>Performance of Linear or Identity Activation Function</figcaption></figure><p>The output of the function is not confined between any range; that is, the value of <em>f(z)</em> can go from which not necessarily a problem as we can proceed into the next or final layer by taking the max value of the neurons that have fired after com. Apart from that, the linear activation function has its set of disadvantages such as:</p><ul><li>We observe that the function’s derivative is a constant. That means there is constant gradient descent occurring since there is no relation to the value of z.</li><li>Our model is not really learning as it does not improve upon the error term, which is the whole point of the neural network.</li><li>Since the activation is linear, nesting in 2 or N number of hidden layers with the same function will have no real effect. The N layers could basically be squashed into one layer.</li></ul><p>We see that this function is not fit to handle complex. So in order to fix this issue, we use non-linear functions to enable our model to learn iteratively.</p><h4>Sigmoid Function (σ)</h4><p>The Sigmoid function takes a value as input and outputs another value between 0 and 1. It is non-linear and easy to work with when constructing a neural network model. The good part about this function is that continuously differentiable over different values of <em>z </em>and has a fixed output range.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QxSV9KmPEoM-zAWd5hHMWw.png" /><figcaption><strong>Fig 2:</strong> Performance of Sigmoid Activation Function</figcaption></figure><p>On observation, we see that the value of <em>f(z)</em> increases but at a very slow rate. The mathematical reason is that as <em>z</em> (on the x-axis) increases, the value of e exponent -z becomes infinitesimally small and the value of <em>f(z)</em> become equals to 1 at some point. In other words, the function is susceptible to the vanishing gradient problem that we discussed in the previous lecture. The high-level problem is that models that utilize the sigmoid activation our slow learners and in the experimentation phase will generate prediction values which have lower accuracy.</p><p>Another issue with this function arises when we have multiple hidden layers in our neural network. All the values we are getting through this activation function are positive and sigmoid churns out values of different magnitudes between 0–1 range so it becomes hard to optimize.</p><p>Disadvantages aside, we do see sigmoid function used especially in binary classification models as part of the output layer to capture the probability ranging from 0 to 1. In most cases, we usually encounter problems that involve multiple classes and for such models, we use the Softmax function (Details of this function are further explained in the <a href="https://towardsdatascience.com/softmax-function-simplified-714068bf8156">previous lecture</a>).</p><h4>Tanh Function</h4><p>The Tanh function is a modified or scaled up version of the sigmoid function. What we saw in Sigmoid was that the value of <em>f(z)</em> is bounded between 0 and 1; however, in the case of Tanh the values are bounded between -1 and 1.</p><p>This is neat in the sense, we are able to get values of different signs, which helps us in establishing which scores to consider in the next layer and which to ignore. But, this function still has the vanishing gradient problem that was seen in the sigmoid function. The model slows down exponentially beyond the +2 and -2 range. The change in gradient is very small except within this narrow range.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1It8846pzYayiC0G_7FIBA.png" /><figcaption><strong>Fig 3:</strong> Performance of Tanh Activation Function</figcaption></figure><p>Last bit about Tanh and Sigmoid, the gradient or the derivative of the Tanh function is steeper as compared to the sigmoid function which we can observe in figure 4 below. Our choice of using Sigmoid or Tanh really depends on the requirement of the gradient for the problem statement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PHO7KDb_7nWPJS7x6wgzWQ.png" /><figcaption><strong>Fig 4:</strong> Comparision of Sigmoid and Tanh Activation Function</figcaption></figure><h4>Rectified Linear Unit Function (ReLU)</h4><p>The Rectified Linear Unit or ReLU for short would be considered the most commonly used activation function in deep learning models. The function simply outputs the value of 0 if it receives any negative input, but for any positive value <em>z, </em>it returns that value back like a linear function. So it can be written as <em>f(z)=max(0,z)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TbZnkZYI5vwOQGUBd6uXAQ.png" /><figcaption><strong>Fig 5:</strong> Performance of ReLU Activation Function</figcaption></figure><p>However, it should be noted that the ReLU function is still non-linear so we are able to backpropagate the errors and have multiple layers of neurons. This function was quickly adopted, as ReLU took care of several problems faced by the Sigmoid and the Tanh:</p><ul><li>The ReLU function has a derivative of 0 over half of its range which spans across all the negative numbers. For positive inputs, the derivative is 1. So we have rectified the ‘vanishing’ gradient problem.</li><li>At a time only a few neurons are activated making the network sparse making it efficient (We will see that sparsity is not always a good thing)</li><li>It is computationally economical compared to Sigmoid and Tanh.</li></ul><p>ReLU has its own set of limitations and disadvantages despite being a better activation function than its other non-linear alternatives:</p><ul><li>The function suffers from the dying ReLU problem — For activations correspondent to values of z&lt; 0, the gradient will be 0 because of which the weights will not get adjusted during the gradient descent in backpropagation. That means, such neurons will stop responding to variations in error/input, so the output network becomes passive due to added sparsity.</li><li>It is best used in between the input and output layers, more specifically within hidden</li></ul><p>As a final note on ReLU, there is a way to counter the dying ReLU problem by using a modified version of the function called the ‘Leaky’ ReLU. To put it briefly, we take the <em>z</em> &lt; 0 values that form the y = 0 line and convert it into a non-horizontal straight line by adding a small, non-zero, constant gradient α (Normally, α=0.01). So our new formula for <em>z</em> &lt; 0 is <em>f(z) = αz</em>.</p><p>For more information, I have attached the Research paper for Leaky ReLU: <a href="https://arxiv.org/abs/1502.01852">https://arxiv.org/abs/1502.01852</a></p><h4>Bonus: Non-linear Cube Activation Function</h4><p>There is one interesting activation function that is applied to Part of Speech (POS) tagging that works better than all the other functions mentioned in the article. The paper by Manning and Danqi Chen (2014) proposed a novel way of learning a neural network classifier for use in a greedy, transition-based dependency parser for natural language processing. They introduce a non-linear cube function denoted by z³ which is further elaborated in the equation of the hidden layer below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/441/0*HCRhhriTPCxKeakD.png" /><figcaption><strong>Fig 6:</strong> Cube Activation function that takes weights of the words, tags, and labels along with its bias.</figcaption></figure><p>The equation above signifies the weighted sum of the input units, which in this case are the word, tag and label embeddings plus the associated bias. The sum is then raised to the power of 3 to compute the output value for that iteration.</p><p>The performance of this particular activation function can be seen in the graph plot below. It is not to scale but we can determine the error terms are reduced as learning is happening at an exponential rate (<em>z³</em>) for this case.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w5TvGMgw1ndNlPEl2doFpg.png" /><figcaption><strong>Fig 7:</strong> Performance of Cube (z³) Activation Function</figcaption></figure><p>The paper is thorough in its experiments where the model is rigorously tested against other state-of-the-art models and to quickly draw a comparison between the activation functions with respect to performance, we observe that:</p><p>Non-Linear Cube &gt; ReLU &gt; Tanh &gt; Sigmoid</p><p>To understand more about how the model runs, I have attached the link of Manning’s paper below.</p><p>A Fast and Accurate Dependency Parser using Neural Networks:<br><a href="https://cs.stanford.edu/~danqi/papers/emnlp2014.pdf">https://cs.stanford.edu/~danqi/papers/emnlp2014.pdf</a></p><h4>Conclusion</h4><p>We learned the numerous activation functions that are used in ML models. For researchers, these functions are used to draw a comparison of what works best given the problem statement. There is no hard and fast rule for selecting a particular activation function. However, it depends upon the model’s architecture, the hyperparameters and the features that we are attempting to capture. Ideally, we utilize the ReLU function on our base models but we can always try out others if we are not able to reach an optimal result.</p><p>As a last note, feel free to comment on your own versions of non-linear activation functions that you have designed, the high-level overview and in what scenario it performs best.</p><p><em>Spread and share knowledge. If this article piqued your interest, give a few claps as it always motivates me to write out more informative content. For more data science and technology related posts follow me </em><a href="https://medium.com/@hamzamahmood"><em>here</em></a><em>.<br>I am also available on </em><a href="https://www.linkedin.com/in/hmahmood93/"><em>Linkedin</em></a><em> and occasionally </em><a href="https://twitter.com/mahmooyo"><em>tweet</em></a><em> stuff as well. :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83ff7f46a6bd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/activation-functions-in-neural-networks-83ff7f46a6bd">Activation Functions in Neural Networks</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Softmax Function, Simplified]]></title>
            <link>https://medium.com/data-science/softmax-function-simplified-714068bf8156?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/714068bf8156</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[naturallanguageprocessing]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Mon, 26 Nov 2018 11:33:05 GMT</pubDate>
            <atom:updated>2018-11-26T23:29:24.102Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SxgCtnoo_9IhL2G0zyvXMg.jpeg" /></figure><h3>The Softmax Function, Simplified</h3><h4>How a regression formula improves accuracy of deep learning models</h4><blockquote><strong>tl;dr </strong>This is my first post of the series dedicated to topics in Machine/Deep Learning and Natural Language Processing. The post discusses Softmax Regression, where we compute the exponential of the input vector in order to normalize the data set into a probabilistic distribution with values that sums to one. Good for multi-dimensional classification instead of binary classification.</blockquote><h4>Prologue</h4><p>I have taken a profound interest in Machine Learning (ML), specifically exploring the field of Natural Language Processing (NLP). Going through research papers and algorithms, the thing that arouses my curiosity over this field; AI in general, is how there always is room for improvement in optimizing experimental results, of how a 100 lines of code can empower industries to automate tasks that otherwise take long, tedious hours to complete. In cases where we have to learn a data-set, then to output meaningful information that is cognizant or aligned with real world expectations and doing that iteratively to achieve more accuracy through incremental changes, to me is absolutely fascinating.</p><p>An insight about my academic journey would be peculiar to most as majority of my courses were related to product engineering and data analytics. However, I wanted to add some extra flavor to this academic experience. In order to do that, I decided to enroll into a Natural Language Processing course offered by the school’s CS department.</p><p>The course did an excellent job in exposing the vast topics there are to learn from and with so much advancements happening in this field, I was super motivated to start writing about the concepts relevant to ML, NLP and Deep Learning. However, there was a challenge associated with it. The concepts studied in and out of class were already mathematically intensive, with chock full of confusing notations and derivations. I wanted to simplify them in such a way that anyone enthusiastic about certain topics could easily understand them and maybe even apply them in research, career, startups or hackathons.</p><p>So within the first few lectures, the professor took us through material that was relevant to <em>Language Modeling</em>; that is<em>, </em>techniques used by intelligent systems to accurately predict a word(s) or sentence(s) in any given text. The content made sense till the professor introduced a particular term called the S<strong>oftmax Regression</strong>. I was overwhelmed by the notation and symbols that came across as French (or more like Latin?) to me. With a few hours of self study and a trusty cup of warm Hazelnut Coffee, the concept made sense soon after.</p><p>To explain it in a more simplified and progressive manner, we will start out with the definition first, then understand the symbols involved and finally see a coded implementation of the function itself.</p><h4>Definition</h4><p>The Softmax regression is a form of logistic regression that normalizes an input value into a vector of values that follows a probability distribution whose total sums up to 1. The output values are between the range [0,1] which is nice because we are able to avoid binary classification and accommodate as many classes or dimensions in our neural network model. This is why softmax is sometimes referred to as a multinomial logistic regression.</p><blockquote>As an aside, another name for Softmax Regression is Maximum Entropy (MaxEnt) Classifier.</blockquote><p>The function is usually used to compute losses that can be expected when training a data set. Known use-cases of softmax regression are in <strong>discriminative</strong> <strong>models</strong> such as <strong>Cross-Entropy</strong> and <strong>Noise Contrastive Estimation</strong>. These are only two among various techniques that attempt to optimize the current training set to increase the likelihood of predicting the correct word or sentence. (We will touch upon the aforementioned techniques in the next few posts, so stay tuned for those.)</p><p>If you look at it from the outset, the definition may sound off as trivial but in the realm of Machine Learning and NLP, this regression function has been useful as a <em>baseline</em> comparator. Researchers who design new solutions have to carry out experimentation keeping the softmax results as a reference. However, it should be noted that softmax is not ideally used as an activation function like <strong>Sigmoid</strong> or <strong>ReLU</strong> (Rectified Linear Units) but rather between layers which may be multiple or just a single one.</p><h4>Notation</h4><p>The classifier function involves some high-level notation which we are going to dive into next. The picture below illustrates how a Softmax function looks like. Let&#39;s try to understand it piece by piece.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eqQuFgXPUP5L6J_vVH19wg.png" /><figcaption>A mathematical representation of the Softmax Regression function</figcaption></figure><ol><li>Given a net input parameter in the form of a <strong>one-hot encoded</strong> matrix <strong>θ, </strong>our objective is to predict if the trained set of features <strong><em>x; </em></strong><em>each </em>with its own set of weights, are a class of <strong><em>j</em></strong>. A one-hot matrix consists of binary values with the number 1 representing an element in the iᵗʰ position of the column while the rest are 0s (a relatively large matrix runs the risk of <em>sparsity</em>, which we will talk about in the next post).</li><li>In the formula we compute the <strong>exponential </strong>of the input parameter and the <strong>sum of exponential parameters </strong>of all existing values in the inputs. Our output for the Softmax function is the ratio of the exponential of the parameter and the sum of exponential parameter.</li><li><strong>θ</strong>, on a high level is the sum of the score of each occurring element in the vector. In a generalized form we say that θ is the transpose of the weights matrix <strong><em>w</em></strong>, multiplied by the feature matrix <strong><em>x.</em></strong></li><li>The term w₀x₀ is the bias that needs to be added on every iteration.</li></ol><h4>Code Implementation</h4><p>Implementing the code is extraordinarily easy and the fun part is that it’s only one line considering we have the necessary Python helper functions at our disposal. In the Github gist below I have used both <strong>Numpy</strong> and <strong>Tensorflow </strong>to write the Softmax function as explained in the previous section. The two mentioned libraries are popularly used to carry out mathematical and neural network related operations.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/a9168e640c0ae4e7c97dd32189c11584/href">https://medium.com/media/a9168e640c0ae4e7c97dd32189c11584/href</a></iframe><h4>Conclusion</h4><p>We shall see further in the next post how we utilize the softmax function when we compute it as part of a log-loss neural network to minimize prediction errors within word embeddings. We will take text samples which will allow us to understand more deeper concepts about NLP and its practical usage in the real world.</p><p>So stay tuned. :)</p><h4>Works Cited</h4><p><a href="http://dataaspirant.com/2017/03/07/difference-between-softmax-function-and-sigmoid-function/">Difference Between Softmax Function and Sigmoid Function - Dataaspirant</a></p><p><a href="https://sebastianraschka.com/faq/docs/softmax_regression.html">https://sebastianraschka.com/faq/docs/softmax_regression.html</a></p><p><em>Spread and share knowledge. This post is the first among many in this series dedicated to understanding core concepts about NLP. If this article piqued your interest, give a few claps as it always motivates me to write out more informative content.Also, do follow my profile for more tech related articles. — Hamza</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=714068bf8156" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/softmax-function-simplified-714068bf8156">Softmax Function, Simplified</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Reaching 25]]></title>
            <link>https://hamzamahmood.medium.com/reaching-25-8730fb0e7458?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/8730fb0e7458</guid>
            <category><![CDATA[mindfulness]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[advice]]></category>
            <category><![CDATA[self-improvement]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Sun, 12 Aug 2018 08:05:33 GMT</pubDate>
            <atom:updated>2018-08-12T20:02:41.130Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7KCIzjE-Ncp8I0rIlvn5mw@2x.jpeg" /></figure><p>Today I turn a quarter of a century old. Am I excited, jubilant, ecstatic or am I simply indifferent?</p><p>Its a mixed feeling to be honest. I am excited yet I am not. Though, there is a peculiar feeling which I could not find the right words to describe it. The spark of being a child, a toddling baby that had not a single worry about life, now a fully matured adult. I may reminisce the joy of my youth from time to time; yet not so much, since being a child meant being dependent on elders.</p><p>To me, being 25 today means I am more wiser than before, goal-driven, empathetic to people, more sensitive to my surroundings, mindful of my actions, appreciative of nature’s beauty, responsible of where and how I spend money and finally pragmatic when making decisions.</p><p>The points below are an aggregation of what I have learned as I tread onwards into adulthood. This may or may not resonate with you but one could take these totally on observational value.</p><ul><li>You cannot push anything to happen. If there is something you expect to happen, it will happen at its own pace. However, it goes without saying that taking <strong>initiative</strong> is key to start something whether it is a business, a relationship, a healthy discussion or a revolution.</li><li>Life is short, even shorter are your relationships with people. The only eternal relationship is the one with your parents. Others may digress but this is what I believe during this stage of my life, an opinion which can ofcourse change with time.</li><li>Using social media is becoming more of a <em>distraction</em> than a useful tool. Coming from a tech person this may be strange but cutting down on social media usage will help you become more productive. I would frequently shift between different apps to check how my friends are doing or how they are spending their lives, instead of investing that time on myself, to become better in my art. It is sort of ingrained in our very Pakistani-oriented minds, the habit of taking interest in other people’s lives. Consequently, I am trying to pursue this mantra of “The less I know the better” due to which I am often finding myself in a more peaceful state than ever before.</li><li>Your mental health is tantamount to your well being. Having thoughts, feelings and concerns dwelling inside you will eat you up. I have tried to battle OCD, anxiety and stress; which I have to confess are equivalent of otherworldly demons that get respawned infinite times in a game, the have made life a literal hell. On a positive note, what has really helped me significantly is to be more precise and straightforward in sharing my opinion or feelings with a person, within a group of people or on a public forum such as Facebook. Assumptions and doubts (ingrained in us thanks to that evil saas or bahu in Indian dramas) are bad habits that need to go.</li><li><em>Negativity</em> is the root of all evil. People who bring you down, or belittle you at any given opportunity are bad for you. However, friends who bash you on your bad habits/aspects, giving you constructive criticism are the ones who mold your character. They practically guide you on the do’s and dont’s of life. Not that all of their advice have to be followed as each individual is flawed in their own manner. Still, I have been lucky enough to have one such friend who has always been the guiding light in several departments.</li><li>When conversing with friends and professionals, we usually ask them about their <strong>future plans. </strong>Don’t be that person! No individual knows where their fate lies, who they are going to be with, what career path they aim to reach at in the next 5 years. By changing the narrative by talking about the present moment will make for a far more engaging conversation.</li><li>You start to realize the frailty of your parents. Start caring for them more, call them if you are living abroad; maybe once or twice a week. After all, they are the ones who taught you how to love in the first place.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/362/1*x-BKM5pnFAB5mYgYdamHAQ.jpeg" /><figcaption>As you can see, the ‘Scooter of Life’ is on auto-pilot and I have the slightest clue of where I am headed</figcaption></figure><ul><li>The most important asset you have is neither time nor money but attention. You may have all the free time in the world and all the money to spend; yes, they are important to maintain a standard of living, but if you’re not experiencing life (interaction with mother nature, human beings, animals…) if you’re not giving attention to those in your surrounding then are you really living? The rhetoric is meant to emphasize on the importance of the bond you have with people. Appreciating the little things has always helped me in staying ‘human’, otherwise I would be totally lost in the void of work and study — which becomes an God awful routine.</li><li>To impart free knowledge to a person is something I have always been a believer of. Having spent hundreds and thousands of currency in education is well and good to land you a job, a meaningful project or assist you in research but should knowledge be something that you keep to yourself? Everyone has a skill or two that they could teach it to someone who can make a living out of it. I know people in my country lack the digital skills which are considered a bare minimum to do a clerical job. Once resources allow me, I wish to someday open a center that offers free schooling to adults on basic digital tool to help them enter the task force.</li><li>Lastly, every human being is going through a stage of evolution. We learn from our mistakes, we make better decisions after a series of wrong ones and we become a better version of ourselves slowly. The key is never to believe that you know everything and to be always open over listen to differing views. We may read a book about a certain topic or hear a talk on a religious topic, we blindly accept this is as the only point-of-view that matters. This stubbornness extends into their personalities as well, having encountered people of such nature. Debate with them, present to them what the other side of the story is and see if there is a shift in mindset. What this entails is the power of communication to change a person’s belief; maybe for the better, to nourish the minds of future generations that is more tolerant and empathetic to differing religions, ideologies and ethnicities.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8730fb0e7458" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Typography Impacts Your Business Outlook — A Primer]]></title>
            <link>https://medium.com/startup-grind/how-typography-impacts-your-business-outlook-a-primer-3f092261d294?source=rss-9a046831034------2</link>
            <guid isPermaLink="false">https://medium.com/p/3f092261d294</guid>
            <category><![CDATA[web-design]]></category>
            <category><![CDATA[business-tips]]></category>
            <category><![CDATA[typography]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Hamza Mahmood]]></dc:creator>
            <pubDate>Sun, 22 Jul 2018 17:02:24 GMT</pubDate>
            <atom:updated>2020-08-30T05:06:32.938Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OdTKwOeXi5JpLKD381kxjw.png" /></figure><h4>For any up and coming business, establishing a proper mode of communication with users is mission critical to their success.</h4><h4>Online Businesses</h4><p>If the business is online, text would be considered as the most basic yet highly effective way to reach their target audience.</p><h4>Designers</h4><p>For designers, typography has real potential to make or break a business. This is a particular area of interest for UI/UX design studios— to build brand identity organically, establish effective communication and streamline content delivery on different platforms through well-thought out typefaces.</p><p>As implied, text is a medium used to inform, discuss, emphasize and encourage new visitors to interact and communicate with the company. However, the way they present themselves through their content creates a long-lasting first impression among users, which could be their last.</p><h4>But where does the term <em>typography</em>, fit into this whole dynamic?</h4><p><em>Typography</em> is basically how text is presented, in the most literal sense. With each letter stylized in a peculiar way, the design pattern remains the same across the rest of the letters.</p><p>Typography is represented by a typeface or more popularly as a ‘font-family’ showcased through titles, headings and paragraphs across a web-page. If the company is professional consulting service, a healthcare provider or an educational platform — the choice and style of words can make a big difference.</p><h3>This write up aims at informing businesses and startups about how typography is a game changer when it comes to first or last impression.</h3><blockquote>Note: All images included in this piece are my own design. They serve the purpose of explaining a particular concept in each section.</blockquote><h4>1.0 Minimize The Number of Font-Families Being Used</h4><p>Never aimlessly add fonts just because they look nice or if its adding some flavor to your web app. It is better to keep the number of font-families at a minimum; such as 2 or max 3. Going beyond 3 will make your web app seem busy and disorganized, almost to the point that it looks amateurish.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EF14H2eQrlyBXdfcC1ZBhw.png" /><figcaption><strong>Fig 1: </strong>Using more than 3 fonts can give the impression of an unprofessional website</figcaption></figure><h4>2.0 What Audience Are You Trying To Reach?</h4><p>When choosing a set of typefaces for a web app, companies need to carry out research on their audiences. This would ideally involve specifying the following:</p><ul><li>Age bracket</li><li>Demographic</li><li>Cultural Affiliation</li><li>Ethnicity</li><li>Gender</li></ul><p>Looking into each factor allows companies to identify the exact pain-points faced by users and consequently formulate their goal and objectives based on the researched user persona.</p><p>Assuming a scenario, you are to design an educational app or a game targeting children between 4–12, chances are you will be using a font which is playful, glittery or maybe has decoration around letters.</p><p>Fig 2a shows one example of a font-family and includes colors on the even numbered letter. The point is to excite children on looking at the sentence and games utilize this technique to make playful learning experiences.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HnEtTKv4dywmUEGhdSxgHw.png" /><figcaption><strong>Fig 2a:</strong> Font-Family: AR Carter</figcaption></figure><p>However, when designing an online mobile-based app for a maturer crowd that may also include senior folk, the font needs to be legible, bold and of relative bigger size as seen in fig 2b.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xfJDvLtIg0_0vHd1QB8pDA.jpeg" /><figcaption><strong>Fig 2b: </strong>We need to design content for different age groups, both young and old.</figcaption></figure><h4>Terminologies</h4><p>In Design Thinking terminologies, to design an app which is wholly customer-centric, a very fundamental practice is to empathize with the users themselves, by listening to them and mapping out a complete flow from point A to B when addressing the problem.</p><p>If businesses open themselves to the customer’s sentiments and woes, they can make the ‘right’ additions or modifications to their existing line of products. Showing empathy tells customers that you care which goes a long way to building brand loyalty.</p><h4>3.0 Complement Your Selected Font-families, Smartly</h4><p>Lets say your business is a news group or blog channel that prioritizes content over other elements, your designers should choose font-families that favors easy readability.</p><p><strong>By rule of thumb:</strong> text should be non-intrusive for the reader and should flow smoothly; hence, we do not want the selected fonts to be seen in juxtaposition to each other. Instead, complementing font-families goes towards improving the reading experience of users.</p><p>For instance, I experimented with a few known Google Fonts, creating combinations of typefaces that made sense and promoted smooth readability in-line.</p><p><strong>In fig 3a:</strong> I used the <em>Montserrat</em> font as my choice for the heading and <em>Merriweather</em> for the paragraph section. The combo can fit well in an online publishing web and mobile site. The mixture of serif and sans-serif font does well in finding the intersection between traditional and contemporary design.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A9a8B8UFtud8Vld8gnhcEA.png" /><figcaption><strong>Fig 3a: </strong>Heading Font — Montserrat &amp; Paragraph Font — Merriweather</figcaption></figure><p>When considering the section in Fig 3b, we observe the minimalist and straightforward nature of the text. I used <em>Fjalla One</em> for the heading text and <em>Noto Sans</em> for the paragraph. Most modern web designs are employing clean format for text and this typeface pair perfectly reflects the trend we see on most landing pages today.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sQrQiwW29McELR9R6He5Gw.png" /><figcaption><strong>Fig 3b:</strong> Heading Font — Fjalla One &amp; Paragraph Font — Noto Sans</figcaption></figure><h4>4.0 Stick With One Typeface, Attain Mastery Before Moving On</h4><p>If your business is just starting out and have novice designer, it is best practice to stick with one typeface before experimenting with others. Modern typefaces are available with different font weights, which can be helpful in certain situations, like in a button, label or a bold heading on the landing page.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zF-ajIW8pMs0dcIYhwULrg.jpeg" /><figcaption><strong>Fig 4: </strong>Various font weights of the Roboto Typeface</figcaption></figure><p>There is no universal standard for weights but category names have been devised as seen in Fig 4 with words that progress from Thin, Light, Regular to Medium, Bold and Black to imply the proportional thickness of the font.</p><p>You also have the option of using Italics and underlined text but those should only be utilized when absolutely necessary.</p><h4>5.0 Pay Attention to Contextual Sizing of Text</h4><p>Once the designer is done with typeface selection, we proceed on to sizing the text. There are a number of ways and scientific tools to determine the perfect size for typefaces.</p><p>For instance, the Modular Scale system gets input in the form of base font-size and ratio, allowing the tool to scale the text to the appropriate sizes by multiplying the two input values. This is just one tool among many to create size guidelines.</p><p>The scaled sizes of the typefaces are then mapped onto the text list that includes different headings, a body text and caption text, as seen in fig 5.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZIEBn03uDMcUXrGUtECqCA.jpeg" /><figcaption><strong>Fig 5: </strong>Scaled sizes (factor = 1.33) of text in Roboto Font</figcaption></figure><h4>However, there are a number of factors that should be considered when applying the sizes on the text:</h4><p><strong>Text that is too small</strong> can cause stress to a user’s eyes, especially when reading it from a mobile device. It is important that text size is legible enough that users can read it comfortably.</p><ul><li><strong>Text that is too large</strong> can be problematic as well. Big sized text can distract users while they perform a task and it calls attention to itself instead of other graphical elements.</li></ul><h4>6.0 Avoid Using Fonts With Cursive/Handwritten Scripts</h4><p>When presenting your online business, it is ideal to use fonts that are clean, simple and most importantly legible. Fonts that involve calligraphy or ones that are fundamentally cursive scripts; although are beautiful to the naked eye, but at times are insanely difficult to read or make sense of. Such scripts break the flow while scanning through the text.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h2EJUt8K4UItHALf0uUIjA.png" /><figcaption><strong>Fig 6: </strong>Hard or easy to read?</figcaption></figure><h4>7.0 Create a Design Guideline for Your Typography</h4><p>In order to standardize design across your brand which includes your website, product line and social media campaign assets, it is imperative to set up a design guideline. One of the important components of this guideline includes typography.</p><p>In this phase, designers can modify attributes of the text such as color palette, effect, weight and size. I used Figma to draft a quick typography guideline in fig 7.</p><p>However, there are other popular tools as well such as Adobe’s Creative Cloud or Sketch (For Apple) that can be used to construct a similar looking guideline like this one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Bt3ubiSyZk0nio09TkyQ4A.png" /><figcaption><strong>Fig 7:</strong> Typography Guideline Created Using Figma</figcaption></figure><h4>Conclusion</h4><p>Typography is not just about improving readability but contributes towards the look and feel of the web or mobile application.</p><p>It takes long hours to master the use of fonts and text placement, which requires designers to experiment, try and test different styles to find the perfect combination.</p><p>Typography is all around us, in both the digital and physical spaces — in books, in monuments, in historical landmarks, in pieces of paper, on our computers and more. I hope this piece serves as a starting point for your journey towards building the right image for your business.</p><h4><strong>Useful Resources</strong></h4><p>Google Fonts — <a href="https://fonts.google.com/">https://fonts.google.com/</a><br>Adobe Typekit — <a href="https://typekit.com/">https://typekit.com/</a><br>Modular Scale — <a href="http://www.modularscale.com/">http://www.modularscale.com</a><br>Typography Inspiration — <a href="https://www.typewolf.com/">https://www.typewolf.com/</a><br>Figma — <a href="https://www.figma.com/">https://www.figma.com/</a><br>Material Design — <a href="https://material.io/guidelines/style/typography.html">https://material.io/guidelines/style/typography.html</a></p><h4>Works Cited:</h4><ol><li><a href="https://medium.freecodecamp.org/typography-can-make-your-design-or-break-it-7be710aadcfe">Typography can make or break your design</a> by Jonathan Z.White<br>2. <a href="https://medium.springboard.com/the-guide-to-getting-typography-right-in-digital-design-bb61214ff3ad">Getting Typography right in the digital Age</a> by Nick Babich</li></ol><figure><a href="https://www.startupgrind.com/conference//#/code-62518medium"><img alt="" src="https://cdn-images-1.medium.com/max/600/1*0sVSBp_k23gLoEzPETtBuQ.png" /></a></figure><figure><a href="https://www.startupgrind.com/start/#utm_source=medium&amp;utm_medium=medium_post&amp;utm_campaign=medium_bottom"><img alt="" src="https://cdn-images-1.medium.com/max/400/1*bXoa4zCny3cktuMZB-Iicw.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3f092261d294" width="1" height="1" alt=""><hr><p><a href="https://medium.com/startup-grind/how-typography-impacts-your-business-outlook-a-primer-3f092261d294">How Typography Impacts Your Business Outlook — A Primer</a> was originally published in <a href="https://medium.com/startup-grind">Startup Grind</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>