You may be wondering how to master Deep Learning and master neural networks. In this page, you’ll find my approach.
I’m Tom, an engineer specializing in Artificial Intelligence. In 2020, I created Inside Machine Learning, the website on which I regularly publish articles on Deep Learning. Out of passion, I take part in competitions in the field – I recently achieved first place in a worldwide contest.
Getting started in Deep Learning isn’t easy. There are plenty of articles on the subject, whether from a purely mathematical angle or from a practical, easy-to-read point of view.
In these lines, you’ll find what I think is the most effective way of mastering Deep Learning. It’s what enabled me to master neural networks.
How to master Deep Learning – The nerds’ method
A nerd is someone who has a passionate and thoughtful interest in computers, networks, and their use. Often this interest dominates their lives to the point where their social skills suffer.
Oxford
In other words, a nerd is someone who spends his days coding algorithms, and his nights reading Wikipedia pages.
When a nerd wants to familiarize himself with a new field, he follows a precise method.
His method is to read up on the whole field in a technical way: study books, read scientific papers, perform mathematical calculations. His aim is to acquire the entire knowledge of the field, so as to be sure to never make a mistake.
How to master Deep Learning
To master Deep Learning, the nerd starts by studying all the mathematical theories associated with the field: cubic equations, harmonic functions, adjoint matrices.
Once he has acquired a solid understanding of mathematical theories, the nerd can finally tackle Deep Learning without fear.
He then explores the concepts of neural networks: stochastic gradient descent, Bayesian optimization, Markov chain Monte Carlo.
The nerd immerses himself in abstract, intangible theories. For him – it’s thrilling!
At the end of his apprenticeship, the nerd will have an in-depth knowledge of all the theories of Deep Learning. He’ll understand the mathematical workings of a neural network, singletons, forward propagation and even why Artificial Intelligence models use the principle of entropy maximization.
However…
Is this the right method?
The nerd method to master Deep Learning is to learn the full range of mathematical theories related to neural networks.
It makes perfect sense to do this! Once you’ve mastered the science of Artificial Intelligence, you’ll feel infallible.
In the French education system, teachers teach students using this method. So, before studying the French Revolution, you have to study all the kings who preceded it. In 6th grade, history lessons even go back to ancient Egypt.
However, I’m not convinced that, in everyday life, this method keeps a normal person engaged in learning.
If I’m interested in the French Revolution and want to know more about it, I don’t want to spend five years learning about all the events leading up to it. I want to find out more – now!
Imagine, you’ve been hearing about “Deep Learning” for several months. You finally decide to get started, but before you can even begin, you’re asked to study a mountain of mathematical theories.
And even when you’re already at a certain level, the first thing you’re introduced to is the concept of “singleton”. Then you’re invited to code it. After that, you’re asked to program the gradient descent function mathematically. And it doesn’t end there! You’ll then need to understand the concept of forward propagation, activation function, loss function, optimizer, etc.
All this requires a considerable investment of time.
Giving up?
When we come across these kinds of articles or videos, we forget half the knowledge we’ve acquired within hours.
We’re bombarded with nebulous theories without even the slightest chance of actually using neural networks.
Even worse, after hours of studying this abstruse content, you may find yourself wanting nothing more than giving up.
This feeling is normal. It comes from the fact that the people who share this kind of information are not normal people.
They’re nerds.
And the method they use is not suited to people with practical brains.
How to master Deep Learning – The practical brain method
What I mean by a “practical brain” is a person who values efficiency over appearance.
The purpose of the practical brain is not to tick all the boxes of the perfect scientist. It’s about making things that work in the real world.
Its method consists in acquiring knowledge, and experimenting with it directly in the field.
The practical brain isn’t afraid of making mistakes. On the contrary, he knows that it is by practicing, failing and rising again that he will gain experience.
How to master Deep Learning
To master Deep Learning, the practical brain starts by reading up online. It finds easy-to-read content that makes neural networks simple to grasp.
Even if this content isn’t scientifically precise and doesn’t detail all the complex mathematical processes behind Deep Learning, the practical brain gets an overview of what a neural network is.
Then he or she moves straight into action by following a tutorial that enables him or her to build his or her own Deep Learning algorithm.
In a day, or two at most, the practical brain has already created his first neural network and obtained concrete results. He certainly hasn’t understood the whole of what he’s just done, but he’s achieved a goal ✔
It’s through practice that we achieve goals, and it’s through achieving goals that we feel accomplished.
Accomplishment
This feeling of accomplishment will keep you engaged throughout your learning process. It’s a reward you receive every time you take action.
Anticipating this reward encourages you to pursue your training.
The feeling of accomplishment is the key to mastering neural networks.
A normal person can’t feel accomplished in his learning with the nerd method, because he doesn’t get concrete results, he ticks boxes.
🎯 You become accomplished when you achieve your goals.
💥 You become accomplished when you take action.
🤝 You become accomplished when you accept mistakes.
The Practical Brain Method gives you the tools to accomplish your goals. It’s now at your disposal. It’s up to you to make the most of it!
If you’d like to find out more about this method and be guided in your Deep Learning apprenticeship, I’ve designed the Action plan to Master Neural networks.
Inside, you’ll find scientific methods and my personal techniques that I use every day in my Deep Learning projects.
If you’re interested, click here – it’s free: