<?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 Özden Özyurt on Medium]]></title>
        <description><![CDATA[Stories by Özden Özyurt on Medium]]></description>
        <link>https://medium.com/@ozdenozyurt2323?source=rss-479c65f4c5a2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*DiSl8YIZOKD3SlqtAoVFaw.jpeg</url>
            <title>Stories by Özden Özyurt on Medium</title>
            <link>https://medium.com/@ozdenozyurt2323?source=rss-479c65f4c5a2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 02:27:30 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ozdenozyurt2323/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[Label Encoding Vs. One Hot Encoding]]></title>
            <link>https://medium.com/@ozdenozyurt2323/label-encoding-vs-one-hot-encoding-8977b27ce5e?source=rss-479c65f4c5a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/8977b27ce5e</guid>
            <dc:creator><![CDATA[Özden Özyurt]]></dc:creator>
            <pubDate>Fri, 02 Dec 2022 17:44:52 GMT</pubDate>
            <atom:updated>2022-12-02T17:44:52.720Z</atom:updated>
            <content:encoded><![CDATA[<p>Firstly, let’s understand what encoding means. Encoding is the process of converting data into a different format. When you convert the temperature readings from Celsius to Fahrenheit or currency from Japanese yen to US dollars, the original values remain the same. They are just represented differently. In the computer world, coding works the same way. The computer converts data from one form to another. It does this to save storage space or to make transmission more efficient.</p><p>I am gonna talk about the two encoders we use during the data preparation phase. Label Encoder and One Hot Encoder are used in a process where categorical variables are converted into a form suitable for machine learning algorithms so that they can be used in predictions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3yi4OT34zgq7Hb-kpH6gbw.png" /></figure><p>LABEL ENCODİNG</p><p>Label Encoding in Python can be achieved using Sklearn Library. Sklearn provides a very efficient tool for encoding the levels of categorical features into numeric values. Label Encoder encode labels with a value between 0 and n_classes-1 where n is the number of distinct labels. If a label repeats it assigns the same value to as assigned earlier.</p><pre>import pandas as pd<br>import numpy as np<br>from sklearn.preprocessing import LabelEncoder# creating initial dataframe<br>bridge_types = (&#39;Arch&#39;,&#39;Beam&#39;,&#39;Truss&#39;,&#39;Cantilever&#39;,&#39;Tied Arch&#39;,&#39;Suspension&#39;,&#39;Cable&#39;)<br>bridge_df = pd.DataFrame(bridge_types, columns=[&#39;Bridge_Types&#39;])# creating instance of labelencoder<br>labelencoder = LabelEncoder()# Assigning numerical values and storing in another column<br>bridge_df[&#39;Bridge_Types_Cat&#39;] = labelencoder.fit_transform(bridge_df[&#39;Bridge_Types&#39;])<br>bridge_df</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/289/0*QJ7gWDCPUjYPWjz4.png" /><figcaption>categorical caolumn and label-encoded column values</figcaption></figure><p>The problem here is since there are different numbers in the same column, the model will misunderstand the data to be in some kind of order, 0 &lt; 1 &lt;2. The model may derive correlations between two variables, but this clearly may not be the scenario in some other data or prediction set. We use One Hot Encoder to overcome this problem.</p><p>Label encoder is used when:</p><ul><li>The number of categories is quite large as one-hot encoding can lead to high memory consumption.</li><li>When the order does not matter in categorical feature.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dqIChGoZQCMkJkhH" /><figcaption>Label Endoing vs. One Hot Encoding</figcaption></figure><p>ONE HOT ENCODİNG</p><p>Though label encoding is straight but it has the disadvantage that the numeric values can be misinterpreted by algorithms as having some sort of hierarchy/order in them. This ordering issue is addressed in another common alternative approach called ‘One-Hot Encoding’. In this strategy, each category value is converted into a new column and assigned a 1 or 0 (notation for true/false) value to the column.</p><pre>import pandas as pd<br>import numpy as np<br>from sklearn.preprocessing import OneHotEncoder# creating instance of one-hot-encoder<br>enc = OneHotEncoder(handle_unknown=&#39;ignore&#39;)# passing bridge-types-cat column (label encoded values of bridge_types)<br>enc_df = pd.DataFrame(enc.fit_transform(bridge_df[[&#39;Bridge_Types_Cat&#39;]]).toarray())# merge with main df bridge_df on key values<br>bridge_df = bridge_df.join(enc_df)<br>bridge_df</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*VrRGecoyMfsxBk-m.png" /><figcaption>one hot-encoded column values</figcaption></figure><p>One Hot encoder is used when:</p><ul><li>When the order does not matter in categorical features</li><li>Categories in a feature are fewer.</li></ul><p>If you found this article useful give it a <strong>clap </strong>and <strong>share </strong>it with your friends.</p><p><strong><em>Thank You :)</em></strong></p><p><strong>&gt; FOR MORE <em>YOU CAN FOLLOW ME ON </em></strong><a href="https://www.linkedin.com/in/%C3%B6zden-%C3%B6zyurt-87b2301a9?lipi=urn%3Ali%3Apage%3Ad_flagship3_profile_view_base_contact_details%3BwnEMJCd9QMyyo5mdXKexmg%3D%3D"><strong><em>LİNLEDLN </em></strong></a><strong><em>ACCOUNT !!!!!!!</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8977b27ce5e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data Visualization]]></title>
            <link>https://medium.com/@ozdenozyurt2323/data-visualization-208dddf8d66c?source=rss-479c65f4c5a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/208dddf8d66c</guid>
            <category><![CDATA[seaborn]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[visualization]]></category>
            <category><![CDATA[matplotlib]]></category>
            <dc:creator><![CDATA[Özden Özyurt]]></dc:creator>
            <pubDate>Fri, 02 Sep 2022 16:55:12 GMT</pubDate>
            <atom:updated>2022-09-08T17:44:36.199Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/884/1*nWz570puAfPngQADg9oHXw.png" /></figure><p><strong><em>What is Data Visualization?</em></strong></p><p>Data visualization is the graphical representation of information and data. By using visual elements like charts, graphs, and maps, data visualization tools provide an accessible way to see and understand trends, outliers, and patterns in data. Additionally, it provides an excellent way for employees or business owners to present data to non-technical audiences without confusion.</p><p>In the world of Big Data, data visualization tools and technologies are essential to analyze massive amounts of information and make data-driven decisions.</p><p><strong><em>Types of data visualizations</em></strong></p><p>The type of chart used in data visualization varies depending on the problem and the situations associated with it. The most important point when visualizing is to determine the chart type. Because the wrong type of graphic may cause nothing to be understood. I have included as much as I can the types of charts that are used both frequently and more specifically.</p><blockquote><strong>1-)General Types of Visualizations:</strong></blockquote><ul><li><strong>Chart:</strong> Information presented in a tabular, graphical form with data displayed along two axes. Can be in the form of a graph, diagram, or map.</li><li><strong>Table:</strong> A set of figures displayed in rows and columns.</li><li><strong>Graph:</strong> A diagram of points, lines, segments, curves, or areas that represents certain variables in comparison to each other, usually along two axes at a right angle.</li><li><strong>Geospatial:</strong> A visualization that shows data in map form using different shapes and colors to show the relationship between pieces of data and specific locations.</li><li><strong>Infographic: </strong>A combination of visuals and words that represent data. Usually uses charts or diagrams.</li><li><strong>Dashboards: </strong>A collection of visualizations and data displayed in one place to help with analyzing and presenting data.</li></ul><blockquote><strong>2-)Specific examples</strong></blockquote><ul><li><strong>Area Map:</strong> A form of geospatial visualization, area maps are used to show specific values set over a map of a country, state, county, or any other geographic location. Two common types of area maps are choropleths and isopleths.</li><li><strong>Bar Chart:</strong> Bar charts represent numerical values compared to each other. The length of the bar represents the value of each variable.</li><li><strong>Box-and-whisker Plots:</strong> These show a selection of ranges (the box) across a set measure (the bar).</li><li><strong>Bullet Graph: </strong>A bar marked against a background to show progress or performance against a goal, denoted by a line on the graph.</li><li><strong>Gantt Chart:</strong> Typically used in project management, Gantt charts are a bar chart depiction of timelines and tasks.</li><li><strong>Heat Map: </strong>A type of geospatial visualization in map form which displays specific data values as different colors (this doesn’t need to be temperatures, but that is a common use).</li><li><strong>Highlight Table:</strong> A form of table that uses color to categorize similar data, allowing the viewer to read it more easily and intuitively.</li><li><strong>Histogram: </strong>A type of bar chart that split a continuous measure into different bins to help analyze the distribution.</li><li><strong>Pie Chart:</strong> A circular chart with triangular segments that shows data as a percentage of a whole.</li><li><strong>Treemap: </strong>A type of chart that shows different, related values in the form of rectangles nested together.</li></ul><h3>Sooooo, which one should we use????</h3><p>To summarize when the following image is examined: Bubble and scatter plots are used to show the relationship, while a histogram chart is mostly used to show the distribution. On the other hand, bar, line and column charts are preferred in comparison representation, while stacked charts are used in composition representation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qtmsP8Ycdi8HNjMeVywySw.png" /><figcaption>Figure 1: Graphic types and usage areas(Designed by Adedamola )</figcaption></figure><p><strong><em>How to apply it ???</em></strong></p><pre><strong><em>#İmport libraries</em><br>import</strong> pandas <strong>as</strong> pd<br><strong>import</strong> numpy <strong>as</strong> np<br><strong>import</strong> matplotlib.pyplot <strong>as</strong> plt<br><strong>import</strong> seaborn <strong>as</strong> sns<br><strong>from</strong> matplotlib <strong>import</strong> pyplot<br><strong>from</strong> datetime <strong>import</strong> datetime<br><strong>from</strong> datetime <strong>import</strong> time, timedelta</pre><pre><strong><em>#Reading the data from Excel file</em><br></strong>data<strong>=</strong>pd<strong>.</strong>read_excel(&quot;Excel_File.xlsx&quot;)</pre><pre><strong><em>#Similar datas changed with unique representation and row with nan values is eliminated</em><br></strong>data_prepared<strong>=</strong>data<strong>.</strong>dropna()</pre><pre><strong><em>#İgnore file warnings</em><br>from</strong> warnings <strong>import</strong> filterwarnings<br>filterwarnings(&quot;ignore&quot;)</pre><pre><strong><em>#Adding a new feature which shows the seasons of the year with respect to datas</em></strong><br>season_dict <strong>=</strong> {&#39;Ocak&#39;: &#39;Kış&#39;,<br>               &#39;Şubat&#39;: &#39;Kış&#39;,<br>               &#39;Mart&#39;: &#39;İlkbahar&#39;, <br>               &#39;Nisan&#39;: &#39;İlkbahar&#39;,<br>               &#39;Mayıs&#39;: &#39;İlkbahar&#39;,<br>               &#39;Haziran&#39;: &#39;Yaz&#39;,<br>               &#39;Temmuz&#39;: &#39;Yaz&#39;,<br>               &#39;Ağustos&#39;: &#39;Yaz&#39;,<br>               &#39;Eylül&#39;: &#39;Sonbahar&#39;,<br>               &#39;Ekim&#39;: &#39;Sonbahar&#39;,<br>               &#39;Kasım&#39;: &#39;Sonbahar&#39;,<br>               &#39;Aralık&#39;: &#39;Kış&#39;}<br>data_prepared[&#39;Mevsim&#39;]<strong>=</strong>data_prepared[&#39;Ay&#39;]<strong>.</strong>apply(<strong>lambda</strong> x: season_dict[x])<br>data_prepared<strong>.</strong>head()</pre><pre><strong><em>#The number of repetitions of categorical datas</em><br></strong>categorical <strong>=</strong> (&quot;Ay&quot;,&quot;Hafta&quot;,&quot;Bölge&quot;,&quot;İlçe&quot;,&quot;İstasyon&quot;,&quot;HücreNo&quot;,&quot;Fider&quot;,&quot;FiderinAçmaNedeni&quot;,&quot;Mevsim&quot;)<br><strong>for</strong> k <strong>in</strong> categorical:<br>    print(&quot;For &quot;,k, &quot;=&quot;,&quot;\n&quot;,data_prepared[k]<strong>.</strong>value_counts(),&quot;\n&quot;)</pre><pre><strong><em>#Bar plot of the categorical variable with frequency</em><br>for</strong> k <strong>in</strong> categorical:<br>    plt<strong>.</strong>figure(figsize<strong>=</strong>(12, 8))<br>    data_prepared[k]<strong>.</strong>value_counts()<strong>.</strong>plot(kind<strong>=</strong>&#39;bar&#39;, color<strong>=</strong>&#39;red&#39;)<br>    plt<strong>.</strong>xlabel(k)<br>    plt<strong>.</strong>ylabel(&quot; Arıza Frequency&quot;)<br>    plt<strong>.</strong>title(&quot;Frequency of power cutting wrt. %s&quot; <strong>%</strong> (k))</pre><pre><strong><em>#Convert the object data type to numeric data</em><br></strong>data_prepared[&#39;L1&#39;] <strong>=</strong> pd<strong>.</strong>to_numeric(data_prepared[&#39;L1&#39;], errors<strong>=</strong>&#39;coerce&#39;)<br>data_prepared[&#39;L2&#39;] <strong>=</strong> pd<strong>.</strong>to_numeric(data_prepared[&#39;L2&#39;], errors<strong>=</strong>&#39;coerce&#39;)<br>data_prepared[&#39;L3&#39;] <strong>=</strong> pd<strong>.</strong>to_numeric(data_prepared[&#39;L3&#39;], errors<strong>=</strong>&#39;coerce&#39;)<br>data_prepared[&#39;TOPRAK&#39;] <strong>=</strong> pd<strong>.</strong>to_numeric(data_prepared[&#39;TOPRAK&#39;], errors<strong>=</strong>&#39;coerce&#39;)</pre><pre><strong><em>#The power cut frequency with respect to current value and day</em><br></strong>variables<strong>=</strong> (&quot;Gün&quot;,&quot;L1&quot;,&quot;L2&quot;,&quot;L3&quot;,&quot;TOPRAK&quot;)<br><strong>for</strong> k <strong>in</strong> variables:<br>    plt<strong>.</strong>figure(figsize<strong>=</strong>(7, 5))<br>    plt<strong>.</strong>hist(data_prepared[k], bins <strong>=</strong> 60 , color<strong>=</strong>&quot;purple&quot;)<br>    plt<strong>.</strong>xlabel(k)<br>    plt<strong>.</strong>ylabel(&quot; Sıklık&quot;)<br>    plt<strong>.</strong>title(&quot;Elektrik Kesintisinin %s Üzerinden Gösterimi&quot; <strong>%</strong> (k))</pre><pre><strong><em># SCATTER PLOT</em><br>for</strong> k <strong>in</strong> variables:<br>    <em>#width = 20</em><br>    <em>#height = 10</em><br>    plt<strong>.</strong>figure(figsize<strong>=</strong>(5, 5))<br>    y<strong>=</strong>data_prepared[&quot;Dakika&quot;]<br>    x<strong>=</strong>data_prepared[k]<br>    plt<strong>.</strong>plot(x,y, color<strong>=</strong>&quot;pink&quot;, linestyle<strong>=</strong>&quot;--&quot;, marker<strong>=</strong>&quot;*&quot;, markersize<strong>=</strong>20, markerfacecolor<strong>=</strong>&quot;grey&quot;)<br>    plt<strong>.</strong>xlabel(k)<br>    plt<strong>.</strong>ylabel(&quot;Dakika&quot;)<br>    plt<strong>.</strong>title(&quot;%s &#39;ın dakika üzerinden incelenmesi&#39;&quot;<strong>%k</strong>)</pre><pre><strong><em>#Evaluation of relations of categorical data and minute </em><br>for</strong> k <strong>in</strong> categorical:<br>    plt<strong>.</strong>figure(figsize<strong>=</strong>(10, 5))<br>    sns<strong>.</strong>barplot(x<strong>=</strong>k,y<strong>=</strong>&quot;Dakika&quot;, data<strong>=</strong>data_prepared )<br>    plt<strong>.</strong>xlabel(k)<br>    plt<strong>.</strong>ylabel(&quot;Dakika&quot;)</pre><p><strong>You can access all codes on the </strong><a href="https://github.com/ozdenozyurt/Visualization.git"><strong>github </strong></a><strong>link or gh repo clone ozdenozyurt/Visualization</strong></p><p><strong>&gt; FOR MORE <em>YOU CAN FOLLOW ME ON </em></strong><a href="https://www.linkedin.com/in/%C3%B6zden-%C3%B6zyurt-87b2301a9?lipi=urn%3Ali%3Apage%3Ad_flagship3_profile_view_base_contact_details%3BwnEMJCd9QMyyo5mdXKexmg%3D%3D"><strong><em>LİNLEDLN </em></strong></a><strong><em>ACCOUNT !!!!!!!</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=208dddf8d66c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[15 Top Python Libraries for Machine Learning]]></title>
            <link>https://medium.com/@ozdenozyurt2323/15-top-python-libraries-for-machine-leatning-b9523e3411ec?source=rss-479c65f4c5a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/b9523e3411ec</guid>
            <dc:creator><![CDATA[Özden Özyurt]]></dc:creator>
            <pubDate>Mon, 18 Jul 2022 21:30:47 GMT</pubDate>
            <atom:updated>2022-07-29T13:03:48.176Z</atom:updated>
            <content:encoded><![CDATA[<ol><li><strong>NumPy</strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*NazZvXW5aqR_MzO5" /></figure><p>NumPy (Numeric Python) is a math library that allows us to do scientific calculations quickly. Numpy is based on numpy arrays. Numpy arrays are similar to python lists, but they are more useful than python lists in terms of speed and functionality. Also, unlike python lists, Numpy arrays must be homogeneous, that is, all elements in the array must be of the same data type.</p><p>Also: <a href="https://numpy.org/">https://numpy.org/</a></p><p><strong>2. Pandas</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*-ZxIJ-j0xRI4TeeE" /></figure><p>After collecting the data, making the data suitable for analysis is very important for data analysis. Data preprocessing and data cleaning is the phase that takes the most time for data scientists. Analyzes are easier and faster by organizing data with the Pandas library. Built on NumPy, Pandas is responsible for preparing datasets and points for machine training. Pandas uses two types of data structures, one-dimensional (serial) and two-dimensional (DataFrame).</p><p>Also: <a href="https://pandas.pydata.org/">https://pandas.pydata.org/</a></p><p><strong>3. SciPy</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*W9wofYZUts-8FSKv" /></figure><p>Scipy (Scientific Python), Python’s many scientific and mathematical calculations needed. Necessary modules and functions for operations are provided. (such as Bessel functions or QR factorization). In this respect, Scipy is one of the modules like “toolboxes” in Matlab. and the functions in each module. The functions in question are in submodules in Scipy. Therefore, without using these submodules, first instead of “Scipy” these submodules are “import” separately should be done.</p><p>SciPy is heavily dependent on NumPy for its array manipulation subroutines and includes all of NumPy’s functionality. SciPy is ideal for image processing and provides basic processing capabilities of non-scientific high-level mathematical functions. It is easy to use and fast to execute. It also includes high-level commands that play a role in data visualization and manipulation.</p><p>Also: <a href="https://scipy.org/">https://scipy.org/</a></p><p><strong>4.Scikit-Learn</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*2TwTVjSwR43DRaog" /></figure><p>Scikit-learn or Sklearn is a Python-based library used to build machine learning models. It has many learning algorithms used for regression, clustering and classification. Thanks to its simple, intuitive and consistent interface, scikit-learn is user-friendly. While the use of scikit-learn is limited as it excels only in data modeling, it allows to allow users to share data as they need.</p><p>Also: <a href="https://scikit-learn.org/stable/">https://scikit-learn.org/stable/</a></p><p><strong>5. TensorFlow</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*2J17-jH1uH2h8oH0" /></figure><p>TensorFlow is a free and open source Python library specializing in differentiable programming. The library provides a collection of tools and resources that help make it easy to build DL and ML models and neural networks. TensorFlow’s architecture and framework are flexible and allow it to run on various computing platforms such as CPU and GPU. However, it does its best when working on a tensor processing unit (TPU).<br>TensorFlow can be used to implement reinforcement learning in ML and DL models and lets you directly visualize your machine learning models with its built-in tools. TensorFlow’s high-level APIs are based on the Keras API standard for defining and training neural networks. Keras provides rapid prototyping, cutting-edge research and manufacturing, all with user-friendly APIs.</p><p>Also: <a href="https://www.tensorflow.org/">https://www.tensorflow.org/</a></p><p><strong>6. Keras</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*mmQQY6PzJHICjcJW" /></figure><p>Keras is an open-source high-level Neural Network library, which is written in Python is capable enough to run on Theano, TensorFlow, or CNTK. It was developed by one of the Google engineers, Francois Chollet. It is made user-friendly, extensible, and modular for facilitating faster experimentation with deep neural networks. It not only supports Convolutional Networks and Recurrent Networks individually but also their combination.</p><p>It cannot handle low-level computations, so it makes use of the Backend library to resolve it. The backend library act as a high-level API wrapper for the low-level API, which lets it run on TensorFlow, CNTK, or Theano.</p><p>Also: <a href="https://keras.io/about/">https://keras.io/about/</a></p><p><strong>7. Theano</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*YfTaCqgW5puBy3ba" /></figure><p>Theano is a Python library that allows us to evaluate mathematical operations, including multidimensional arrays, very efficiently. It is mostly used in making Deep Learning Projects. It runs faster on the Graphics Processing Unit (GPU) rather than the CPU. Theano achieves high speeds for problems with large amounts of data. In some cases it can take advantage of GPUs that make it significantly outperform C on the CPU by orders of magnitude.<br>It knows how to take structures and convert them into very efficient code that uses numpy and some native libraries. It is designed to handle the types of computations required for large neural network algorithms often used in Deep Learning.</p><p><strong>8. PyTorch</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*SjQjnjXrxGm_ACLG" /></figure><p>PyTorch(Python Torch) is the Python library for building deep learning models that use the power of graphics processing units. It is to build deep neural networks in tensor computations and tape-based autograde systems with powerful GPU acceleration support. One of the main reasons behind the success of PyTorch is that it is purely Pythonic and can build neural network models without any problems.<br>PyTorch, which uses graphics processing units in projects, is very popular today with the flexibility and speed it provides due to its structure.<br>Being pythonic, with a simple interface and computational graphics, it is TensorFlow’s biggest competitor to date and is currently a highly preferred library in research communities.</p><p>Also: <a href="https://pytorch.org/tutorials/">https://pytorch.org/tutorials/</a></p><p><strong>9. Matplotlib</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*OjHK46hnubhz3bJP" /></figure><p>Matplotlib(Mathematical Plot Library ) library is a numerical plotting library for python and the numpy library. These library outputs work in harmony with GUI (Graphical Interface) plugins such as Tkinter, Qt, and provide object-oriented API to applications. It gives 2 or 3 dimensional visual outputs of numerical mathematical calculations for Data Science studies. It is highly preferred because its documentation content is rich, easy to learn and simple to implement.</p><p>Also: <a href="https://matplotlib.org/">https://matplotlib.org/</a></p><p><strong>10. Seaborn</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*4iR0JxApnh_QZy8u" /></figure><p>Seaborn is a data visualization library built on matplotlib and closely integrated with pandas data structures in Python. Its content is much more comprehensive than matplotlib. Seaborn functions: API for dataset to determine relationship between variables. Automatic estimation and plotting of linear regression graphs. Supports high-level abstractions for multi-point grids. Visualizing univariate and bivariate distribution. Using Seaborn we can draw a wide variety of charts such as: Scatter Charts, Pie Chart and Bar Chart, Scatter Charts, Pair Charts and Heatmaps</p><p>Also: <a href="https://seaborn.pydata.org/">https://seaborn.pydata.org/</a></p><p><strong>11. OpenCV</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/389/0*vBcPSOojEHuJo7qW" /></figure><p>OpenCV (Open Source Computer Vision) contains more than 2500 algorithms for image processing and machine learning. With these algorithms, operations such as face recognition, object discrimination, human motion detection, object classification, license plate recognition, processing on three-dimensional images, image comparison, optical character identification OCR (Optical Character Recognition) can be easily performed.</p><p>Also: <a href="https://opencv.org/">https://opencv.org/</a></p><p><strong>12. NLTK</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/150/0*kG9ygEvHpWsEpuHc" /></figure><p>NLTK (Naturel Language Toolkit ) is a library that provides computer and human interaction, enables the computer to understand human language, and has many different layers for language studies. For example, it allows us to do many projects such as sentiment analysis, text summarization system, text similarity calculation, virtual assistants, translation systems.</p><p>Also: <a href="https://www.nltk.org/">https://www.nltk.org/</a></p><p><strong>13. Caffe</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/140/0*G-PRdeTSgBmgL8KO" /></figure><p>Caffe deep learning structure is designed to be fast and modular. Berkeley Vision and Learning Center — Developed by the BVLC (Berkeley Vision and Learning Center) and its user community. Prepared by Yangqing Jia during her PhD at UC Berkeley. Expression Structure encourages the introduction of new and different ideas or practices. The extensible code structure supports active development. Caffe can process over 60 Million images in a day with a single NVIDIA K40 GPU* processor. This means 1 ms/picture for inference and 4 ms/picture for learning. We believe that Caffe is the fastest convnet application available.</p><p>Also: <a href="https://caffe.berkeleyvision.org/">https://caffe.berkeleyvision.org/</a></p><p><strong>14. Gensim</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*rF_DsQGV7wPrahya" /></figure><p>Gensim is a Python library with tools for vector space modeling and subject modeling. For example, you can use it to capture the semantic integrity between 2 newspaper news. It has an incredible memory usage optimization and processing speed obtained with the help of the numpy library. It is also frequently used for unsupervised learning.</p><p>Also: <a href="https://radimrehurek.com/gensim/">https://radimrehurek.com/gensim/</a></p><p><strong>15. SpaCy</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*3o8P5gjJg6HLSlf4" /></figure><p>SpaCy is one of the latest technologies in the natural language processing world. It is very fast because it is written in Cython language, which depends on Python but has a syntax similar to the C programming language, which is designed to be faster than Python. For this reason, it allows you to work with large volumes of texts. SpaCy is currently the fastest syntax parser on the market. It supports tokenization in 49 languages. This library, which is generally used in name entity recognition (NER) projects, can work integrated with deep learning models.</p><p>Also: <a href="https://spacy.io/">https://spacy.io/</a></p><p>NOTE !! The above libraries can be installed with the help of the following command. — —pip install library</p><p><strong><em>YOU CAN FOLLOW ME ON </em></strong><a href="https://www.linkedin.com/in/%C3%B6zden-%C3%B6zyurt-87b2301a9?lipi=urn%3Ali%3Apage%3Ad_flagship3_profile_view_base_contact_details%3BwnEMJCd9QMyyo5mdXKexmg%3D%3D"><strong><em>LİNLEDLN </em></strong></a><strong><em>ACCOUNT !!!!!!!</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b9523e3411ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Hyper-parameters used in Deep Learning]]></title>
            <link>https://medium.com/@ozdenozyurt2323/hyper-parameters-used-in-deep-learning-633694851f5f?source=rss-479c65f4c5a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/633694851f5f</guid>
            <dc:creator><![CDATA[Özden Özyurt]]></dc:creator>
            <pubDate>Tue, 17 May 2022 13:00:36 GMT</pubDate>
            <atom:updated>2022-06-09T19:37:55.674Z</atom:updated>
            <content:encoded><![CDATA[<p><em>When creating an artificial intelligence model that learns from data, we need some parameters to decide which algorithm or techniques to use in the model. For example, the programmer decides what the k value will be in the KNN classification algorithm. Likewise, the programmer decides which kernel function to use in the SVM algorithm. It is not clear at the beginning of the problem on what basis these parameters will be determined. It varies depending on the data set, the problem or the area in which it will be used. After these factors are finalized, the programmer determines the parameters to be used in the problem in the most appropriate range.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/588/0*RgdUf0WrPN02iCRi.png" /><figcaption>Figure 1: Parameter and Hyper parameter in Deep Learning Applications (Image Source: <a href="http://www.xlgps.com/article/317732.html">http://www.xlgps.com/article/317732.html</a>)</figcaption></figure><ol><li><strong><em>Size and Variety of Data Set</em></strong></li></ol><p>We all know that the size and d iversity of the data set are the most important parameters for learning in deep learning applications. The larger our dataset, the better the learning will be. Not only is the size of the data set sufficient for a good model, its diversity is also important. As the diversity increases, the performance of the model will also increase.</p><p>If we have a very small data set, it is not very suitable to solve the problem with deep learning. The data set should be increased with synthetic data or feature transfer should be done with “transfer learning” methods. Or models such as SVM should be used, which are ideal for solving non-linear problems that work better on small data. The data set can be augmented by synthetic data generation techniques (data augmentation). This method will increase performance to a certain extent.</p><p><strong><em>2. “Mini-Batch” Size</em></strong></p><p>In deep learning applications, learning by processing all data in the data set in the same time is a costly process in terms of time and memory management. Because, in each iteration of learning, gradient descent is calculated on the neural network with the backpropagation process and the weight values are updated in this way. In the calculation process, the number of data and the calculation process increase in direct proportion. As a solution to this problem, data sets are divided into small groups and the learning process is carried out over these groups. The process of processing more than one input value in parts in this way is called “mini-batch”. The value assigned as the mini-batch parameter in the model design; decides how many data the model will process at the same time. In the studies carried out, it has been determined that the loss value increases but time is gained in the case of processing the data in groups (mini-batch).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1014/0*HobWUd6hT_uH3-1G.png" /><figcaption>Figure 2: Error value change when batch and mini-bath are applied (Image credit: Steven Schmatz)</figcaption></figure><p><strong><em>3. Learning Rate and Momentum Coefficient</em></strong></p><p>Updating the parameters in deep learning is done by backpropagation. In the backpropagation process, this update is done by finding the difference by making a backward derivative called “chain rule” and multiplying the difference value with the “learning rate” parameter, subtracting the result from the weight values ​​and calculating the new weight value. The “learning rate” parameter used during this process can be determined as a fixed value, or it can be determined as a step-by-step value (for example, 0.001 until a certain learning step, 0.01 after that step), depending on the momentum value, or learning by adaptive algorithms. can be learned during</p><p>The noise generating methods such as Schoastic gradient descent are normalized with techniques such as exponential weight average and their oscillations are reduced. This normalization process is done with momentum beta coefficients. In this technique, instead of taking the newly produced value as it is, the new value is calculated by adding the previous value at the rate of the beta coefficient. Thus, a faster method is created by reducing noise and oscillations in the graph.</p><p><strong><em>4. Number of Epochs</em></strong></p><p>While the model is being trained, not all of the data are included in the training at the same time. They take part in education in a certain number of parts. The first piece is trained, the performance of the model is tested, and the weights are updated according to the success with backpropagation. Then the model is retrained with the new training set and the weights are updated again. This process is repeated at each training step to try to calculate the most appropriate weight values ​​for the model. Each of these training steps is called an “epoch”.</p><p>Since the most suitable weight values ​​to solve the problem in deep learning are calculated step by step, the performance will be low in the first epochs, and the performance will increase as the number of epochs increases. However, after a certain step, the learning situation of our model will decrease considerably.</p><p>The model usually takes a long time to train; There are models that take days or months to train. This is common in deep learning. For this reason, it is tried to shorten the training process as much as possible with other hyper parameters.</p><p>The size of the epoch number also varies according to the problem type. For example, the number of epochs should be kept larger in pattern-learned RNNs (Recurrent Neural Networks) compared to other models. As the number of epochs increases, the performance of the model increases significantly. Training can be terminated at these points, as performance will increase in very small units after a certain epoch.</p><p><strong><em>5. Determination of Weight Initial Values</em></strong></p><p>How the weights are determined affects the learning and speed of the model.<br>If the W weights are initialized to 0 at the beginning of the model, the input values appear as the output, as the matrix product is a sum. Since y = f(x,w) -&gt; w.x, since the values of the w matrix are 0, the addition of x and the matrix containing zero will still give x itself. Therefore w weights should not be initialized to 0 initially.</p><p>If the W weights are started from randomly small numbers at the beginning of the model, the model will work in small networks, but will result in a heterogeneous distribution of activation between network layers.</p><p><strong><em>6.Dilution (Dropout) Value and Layers to be Diluted</em></strong></p><p>It has been shown that dilution of nodes below a certain threshold value in fully connected layers increases performance. In other words, forgetting weak information contributes positively to learning.</p><p>The dropout value is generally 0.5. It varies according to the problem and the data set. When the dropout value is used as the threshold value, it is defined as a value in the range of [0,1]. Finally, it is not mandatory to use the same dropout value in all layers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/460/0*s5zkty812wH_nvtZ.gif" /><figcaption>Figure 3: Example dropout application (Image credit: Cristina Scheau)</figcaption></figure><p><strong><em>7.Convolutional Neural Network (CNN) Kernel Size</em></strong></p><p>In convolutional neural networks, kernels that operate on the matrix are used in each layer. These are like simple Gabor filters. The size of these kernels is highly effective on learning. Because the kernel size and how wide the data will affect each other are decided. Generally, 3x3, 5x5, 7x7 kernels are used. Using a large kernel will cause the image to be smaller after convolution is applied. In this case, small sized kernel such as 3x3 is generally used because it causes loss of information. In the edge detection process, in order to be able to look to the right, left, top and bottom of the pixel being processed, filters that can be centrally composed of odd numbers are used.</p><p><strong><em>8.Selection of the optimization algorithm</em></strong></p><p>In deep learning applications, the learning process is essentially an optimization problem. Optimization methods are used to find the optimum value in the solution of nonlinear problems. Used as a publication in deep learning; Optimization algorithms such as stochastic gradient descent, adagrad, adadelta, adam, RMSProp, adamax are used. There are differences between these algorithms in terms of performance and speed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/335/0*KPB3EMcm2lxVeG35.png" /><figcaption>Figure 4: Runtime graph of major optimization algorithms on MNIST dataset</figcaption></figure><p><strong><em>9. Activation Function</em></strong></p><p>Activation functions are used for non-linear transformation operations in multilayer artificial neural networks.<br>Activation functions add non-linearity to the model. In our linear function y = f(x,w) in the hidden layers, matrix multiplication is performed and the weight of the neurons is calculated, and the output is converted to a non-linear value. Because deep learning methods are more effective than other methods in solving problems with nonlinear structure, the problem that is tried to be solved by deep learning methods is generally a nonlinear non-linear problem. The conversion of the value obtained as a result of matrix multiplication into non-linear is done with activation functions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/770/0*VOoId7dPZWO5heTU.png" /><figcaption>Figure 5: Graphs of the most known activation functions</figcaption></figure><p><strong>Conclusion;</strong></p><p>1)The more the data set and its diversity, the greater the learning.<br>2) The more groups there are, the higher the loss value, but time is gained.<br>3) A high learning rate causes oscillation. Being small too<br>This will cause learning to take too long.<br>4) As the number of epochs increases, so does the performance.<br>5) Using large kernels may cause information loss.</p><p>The parameters described in the article are the most frequently encountered hyperparameters in deep learning applications. In addition to these described parameters, there are also parameters for different network structures.</p><p>If you want to follow me on Linledln account: <a href="https://www.linkedin.com/in/%C3%B6zden-%C3%B6zyurt-87b2301a9">linkedin.com/in/özden-özyurt-87b2301a9</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=633694851f5f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Artificial Intelligence, Machine Learning VS. Deep Learning]]></title>
            <link>https://medium.com/@ozdenozyurt2323/artificial-intelligence-machine-learning-vs-deep-learning-29a8ff8dc0b4?source=rss-479c65f4c5a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/29a8ff8dc0b4</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[ml-vs-dl]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Özden Özyurt]]></dc:creator>
            <pubDate>Tue, 19 Apr 2022 22:10:51 GMT</pubDate>
            <atom:updated>2022-04-20T22:21:59.618Z</atom:updated>
            <content:encoded><![CDATA[<h4>What is Artificial İntelligence?</h4><p>Computer science and ‘artificial intelligence’ concepts were born thanks to electromechanical devices produced with Crypto analysis requirements during World War II in 1943. In general, AI is the broadest term used to classify machines that mimic human intelligence. According to Alan Turing’s definition fall under the category of AI “systems that act like humans.” There are 3 main categories of AI: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI) and Artificial Super Intelligence (ASI)</p><p>In its simplest form, artificial intelligence is a field that combines computer science and robust datasets to enable problem solving. It covers the sub-fields of machine learning and deep learning, which are often mentioned together with artificial intelligence. These subdomains consist of AI algorithms that make predictions or classifications based on input data, with it trying to build systems. Today, it includes artificial intelligence algorithms from the electronics in our house to the clothes we wear.</p><p>Sub-field of AI:</p><p>&gt;Machine Learning</p><p>&gt;Deep Learning</p><p>&gt;Natural Language Processing</p><p>&gt;Neural Networks</p><p>&gt;Speech Processing</p><p>&gt;Computer Vision</p><p>&gt;Evolutionary Computation</p><p>&gt;Expert Systems</p><p>&gt;Pattern Recognition</p><p>&gt;Genetic Programming</p><p>&gt;Fuzzy Logic</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/353/1*bgVQWes43iyVJcD9pGBs4A.png" /></figure><h4><strong>Machine Learning VS. Deep Learning</strong></h4><p>As mentioned above, both deep learning and machine learning are sub-fields of artificial intelligence. Simply, Deep learning is actually a sub-field of machine learning. İt is actually comprised of neural networks. “Deep” in deep learning refers to a neural network comprised of more than three layers. Deep learning structures algorithms like CNN, RNN, GAN and DBN in layers to create an “artificial neural network” that can learn and make intelligent decisions on its own.<br>Machine learning uses algorithms to parse data, learn from that data, and make informed decisions based on what it has learned. There are four types of machine learning algorithms: supervised, semi-supervised, unsupervised and reinforcement.</p><p>Machine learning and deep learning have different pros and cons according to their application areas. For example, while DL can automatically discover the features to be used for classification, ML requires these features to be provided manually. You can see a clear comparison of this in the table below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/607/1*9bX4SCNNSassj7IAegOi7w.png" /><figcaption>Parameter comparison table of ML and DL</figcaption></figure><p>In summary, machine learning and deep learning have different places where they excel one another. The aim here is to determine the most suitable method for the problem and to produce a solution accordingly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/733/0*EVkjg67uFexGucpN" /><figcaption>Visually the difference between ML and DL.</figcaption></figure><p>If you want to follow me on Linledln account: <a href="https://www.linkedin.com/in/%C3%B6zden-%C3%B6zyurt-87b2301a9">linkedin.com/in/özden-özyurt-87b2301a9</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=29a8ff8dc0b4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>