Why Deep Learning is everywhere

And why does so much of Data Science and AI revolve around Neural Networks?

Deep Learning is everywhere.

But not many people can actually explain what it is. How is Deep Learning different from Machine Learning? Or why Neural Networks became the poster child for AI even though we had so many kinds of architectures. What is the mathematical reason that AI researchers scale up Neural Networks but not Logistic Regression?

Today, we will be covering that. Even if you’re not into Machine Learning, this will help you understand the background behind one of the most important fields in our society.

A preview of what is to come. Don’t worry, I won’t use so many complex words.

Important Details

1. What is Machine Learning- To those of you that absolutely know nothing about this field, here is an ELI5 explanation- In Machine Learning, we feed a lot of data to the model and have it try to spot patterns. The feeding of the data is called ‘training’. In a more mathematical sense, during our model training our model tries to find a function to match the data given to it.
2. What is Deep Learning- Deep Learning is a subset of Machine Learning. In it, we use Neural Networks (which are basically weighted graphs) with lots of layers. However, there is a catch. Simple neural networks (even very big ones) still produce linear functions. This is true for most Machine Learning methods. However, turns out that introducing non-linearity is a game changer.
3. Wait….What is Linearity? How can a graph be Linear?- Here, I’m using the mathematical definition of a linear function. In simple words, a function is linear if you it meets two conditions-

Image Source

Linearity is important enough as a concept that I will do a dedicated Monday post on it. It’s a concept you will see show up a lot, in all avenues of statistics and engineering.

4. So what does this have to do with Deep Networks- Turns out that we can approximate any continuous function by taking a combination of non-linear functions. I’m highlighting continuous b/c many people misunderstand this theorem to assume that you can approximate any function. This faulty assumption has wasted a lot of time, energy, and VC money. Very big deep networks work because they are able to try a lot of non-linear functions and all kinds of combinations of them.

To any of you that are interested in the Mathematical proof, read this.

5. Why everyone is using Deep Learning- The reason we see them everywhere is that humans are attracted to shiny new objects. And we’re lazy. If I’m given a data-analysis problem, I can use Deep Learning to get some kind of a result, so people often resort to them without thinking too much. And since it’s the trendy thing these days, no questions about whether Deep Learning makes sense. The whole ethos of my Machine Learning content is to highlight that you can do lots of things to improve your data analysis that don’t involve more training data, bigger Deep networks, and more training.

This should be a reasonably good overview of why Deep Learning seems to have taken over much of AI these days. If you’re a decision maker then knowing this is crucial to understanding this new field. And if you’re a developer (especially if you’re into Data Analysis), then understanding the universal approximation is crucial to understanding why you see Neural Networks everywhere.

Photo by Josh Riemer on Unsplash

For a wonderful visualization of this concept (Universal Function Approximation), check out the following video. As a creator, I know how hard it can to talk about higher-level Deep Learning ideas without getting into the weeds. Emergent Garden threads this needle brilliantly. I’m not affiliated with his channel in any way, I just think he makes exceptional content. If you are into higher-level AI, add his channel to your list.

For more such writeups, check out my newsletter Technology Made Simple. Tech Made Simple is the best resource for people looking to build amazing careers in Tech. It will help you conceptualize, build, and optimize your solutions. It covers everything from technical aspects from System Design, Computer Science concepts, and Leetcode Problem solving to detailed guides on networking and building your career. Save your time, effort, and money by finding all your needs met in one place. Use the link here to get 20% off, for up to a whole year.

I created Technology Made Simple using new techniques discovered through tutoring multiple people into top tech firms. The newsletter is designed to help you succeed, saving you from hours wasted on the Leetcode grind. I have a 100% satisfaction policy, so you can try it out at no risk to you. You can read the FAQs and find out more here

Feel free to reach out if you have any interesting jobs/projects/ideas for me as well. Always happy to hear you out.

Reach out to me

Use the links below to check out my other content, learn more about tutoring, or just to say hi. Also, check out the free Robinhood referral link. We both get a free stock (you don’t have to put any money), and there is no risk to you. So not using it is just losing free money.

To help me understand you fill out this survey (anonymous)

Check out my other articles on Medium. : https://rb.gy/zn1aiu

My YouTube: https://rb.gy/88iwdd

Reach out to me on LinkedIn. Let’s connect: https://rb.gy/m5ok2y

My Instagram: https://rb.gy/gmvuy9

My Twitter: https://twitter.com/Machine01776819

If you’re looking to build a career in tech: https://codinginterviewsmadesimple.substack.com/

Get a free stock on Robinhood: https://join.robinhood.com/fnud75