<?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 Prajogo Atmaja on Medium]]></title>
        <description><![CDATA[Stories by Prajogo Atmaja on Medium]]></description>
        <link>https://medium.com/@prajogo.atmaja?source=rss-58187e4d2d38------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*mBHWfKzbfw1G-ddH</url>
            <title>Stories by Prajogo Atmaja on Medium</title>
            <link>https://medium.com/@prajogo.atmaja?source=rss-58187e4d2d38------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 23:29:25 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@prajogo.atmaja/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[Start Your Deep Learning Journey Here]]></title>
            <link>https://medium.com/@prajogo.atmaja/start-your-deep-learning-journey-here-2502119dd96c?source=rss-58187e4d2d38------2</link>
            <guid isPermaLink="false">https://medium.com/p/2502119dd96c</guid>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Prajogo Atmaja]]></dc:creator>
            <pubDate>Sun, 17 Jun 2018 18:07:25 GMT</pubDate>
            <atom:updated>2018-06-17T18:07:25.911Z</atom:updated>
            <content:encoded><![CDATA[<h3>Short Introduction</h3><p>Starting in a deep learning path seems to be discouraging. In fact most people ended up not trying it because it looks so complicated. However I would say that it is not and today I would like to guide you to start your deep learning journey here.</p><h3>AI, Machine Learning, Deep Learning, what’s all of this about?</h3><p>AI has a very broad term but to generalize it AI is basically a machine’s or software’s intelligence to do some tasks. The word “artificial” itself came because we don’t actually understand what does “intelligence” means. We can’t actually measure it as IQ test is actually inaccurate in my opinion. One’s brain is different from others and it’s so complex that (I won’t explain much about this part). In AI, we could measure the “intelligence” based on its performance upon clearing one’s task.</p><p>Machine Learning is a subset of AI where we approach the problem by let the software or machine to “learn” by itself, without having to program it explicitly. Basically, you tell your PC “Hey, I have tons of data. learn this well and tell me what’s will the outcome be if I give this kind of data (This is called supervised learning by the way).”</p><p>Deep Learning is a subset of Machine Learning where we literally have a “deep” hidden layer in our model, so that our model is pretty complex. One example would be an object recognizer, where the model has a lot of layers in the model to actually recognize a thing.</p><h3>Why Now?</h3><p>Actually, Machine learning is not something new. It has been discovered since 1950’s. Artificial intelligence came in a form of artificial neural network or ANN to be short. ANN simulates the work of a single neuron in our body. Back then, ANN was actually a big thing. There are a lot of researchers went into those area. I’m not an expert in history but however, as far as I know, ANN researches stumbled because they can’t actually solve an XOR problem in ANN. My professor said that many researchers frustrated and ANN research went dead. Years later it was found that to solve an XOR problem, you need another layer for it to make sense.</p><p>Nowadays, Deep learning are far more approachable since computing power has grown very large, and accessible to researchers, and also home users. In fact, the number of enthusiasts and developers helps the community to grow. Also, it is easier for us to gather data now.</p><h3>Out of All Technologies, Why Deep Learning?</h3><p>Deep learning is pretty much “black magic”. You can’t actually explain why does it work, but it worked! Deep learning is now a pretty hot topic throughout the technology industries and they will pay those who masters it well!</p><h3>I’m not From A Computer Science Major, Will I Be Able to Do It?</h3><p>Absolutely! What you actually need is a very basic programming skill and some math! From there, you can actually build your own model, but perhaps it will be very long to train. You can actually read some books, or join some online classes which will tell you more about vectorization, dropout, and much more.</p><p>One book that I would recommend you to read is a book by Aaron Courville, Ian Goodfellow, and Yoshua Bengio which I would put the link <a href="http://www.deeplearningbook.org/">here</a>. Don’t get discouraged with the symbols and such. Computing is all about logic. Read well, visualize in your brain and I’m pretty sure you can get it.</p><p>If you are a audio visual type of person, I would recommend you to take an online class. There are some several online classes out there but I would recommend Andrew Ng’s Deep Learning class from Coursera, link <a href="https://www.coursera.org/specializations/deep-learning">here</a>. It is a paid class but he actually explains very well and worth the price. You can also ask for a financial aid if you needed. The drawback of this class is basically it’s a semi one-direction talk. Because it’s a video but you could ask questions on the forum if you need. Another drawback is the video is like when you take your class, the way he explains is somewhat professor-like and I know some of you wouldn’t like it. Another great guy would be Siraj Raval. He creates some contents on youtube and also a pretty interesting guy, links <a href="https://www.youtube.com/channel/UCWN3xxRkmTPmbKwht9FuE5A">here</a>. If I’m not wrong, he also create a paid class but I’m not sure. I wouldn’t say that one is better than the other but it comes back to preferences whether you like a professor-like class or some friendly guy explains you something.</p><h3>Let’s Start then, What do I Need?</h3><p>You don’t actually need those fancy workstations to actually start your deep learning journey, as you are not dealing with hundred gigs of data. Sure it would be long but well it’s a start. later, as you get hang on the basics and you’re thinking of advancing, I would suggest you to build a decent workstation.</p><p>You will just need your PC or Laptop. Ubuntu installed is preferable but that doesn’t mean you can’t do it in Windows or Mac OS. Then set up a python environment in your PC. I would recommend using <a href="https://anaconda.org/">anaconda</a> to start with. I would leave the installation guide to Hussnain Fareed’s <a href="https://medium.com/@hussnainfareed/setup-an-environment-for-machine-learning-and-deep-learning-with-anaconda-in-windows-5d7134a3db10">post</a>. After you’re done, I would suggest to try some tutorials from <a href="https://www.tensorflow.org/tutorials/">tensorflow</a>. Learn by doing is the best way to get on something. When you get a hang to it, start to explore kaggle to test your skill! Start by doing their deep learning tutorial track <a href="https://www.kaggle.com/learn/deep-learning">here</a>. Finally, try to compete with others there.</p><h3>What’s Next?</h3><p>The next thing you do would be getting a better PC to do the job. You wouldn’t want to train just a hundred meg of data for hours. Why do I suggest building a PC? Laptops aren’t designed to be working on a full load 100% for a long time (24h+) and cloud services are in my opinion expensive. I don’t think that I would pay 10 dollars an hour to train my model. That’s why I would recommend you to build one.</p><p>Building a PC really depends on your budget but do allocate your budget on graphics card as deep learning computes very much on GPU rather than CPU. As for GPU, I would suggest taking NVIDIA’s GPU to support its CUDA. It’s not that AMD’s GPUs are bad but AMD hasn’t started to support its GPU fully on deep learning (but they will in the future). If you are on a low budget, GTX 1050Ti will do the job. If you have the money, go for a GTX 1080Ti, 2 if possible. I wouldn’t go for Titan X or Titan Xp as the price doesn’t add up in my opinion.</p><p>The second most important thing would be storage (why?). Computer’s speed is determined by the slowest part of the PC and storage is one. I would suggest to get an M.2 NVME SSD if possible for your operating system since M.2 NVME is plugged directly to the PCIe lane(it’s the fastest lane). SATA SSD will be slower than M.2 NVME since its speed are limited from the SATA connection. I don’t need to explain about HDD, but you need to get one to store your data.</p><p>Next CPU, Motherboard and RAM. Choose which brand are you going to use whether Intel or AMD. I’d choose AMD as it’s less pricey and it works very well. Pick not too expensive, as you most of your computing would be focused on GPU, but still decent, so that your CPU wouldn’t bottleneck your GPU. For motherboard, if you are going for the M.2 NVME SSD then go for motherboards that support M.2. Otherwise boards wouldn’t be a problem unless you are going for a high budget which you would like to upgrade your PC. Make sure that you’ve picked the same CPU socket with your motherboard. RAM, as a rule of thumb, get at least 2x of your GPU memory. I would recommend to get at least 2133 MHz RAM. Make sure to check whether it’s DDR3 or DDR4.</p><p>Power supply would be the last important thing. The rule of thumb is to count your wattage first, then add 100 watt to it. Get a known brand and at least have 80+ bronze(at very least). Others like casing, cooling, etc., I will give it back to you.</p><p>With your newly built workstation, it will definitely boost up your productivity</p><h3>Conclusion</h3><p>I hope that with this, you will be able to take your first step into deep learning and have fun with it. Cheers!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2502119dd96c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>