Entering the World of Machine Learning: Index 0

What is machine learning? That was the question I faced several weeks ago.

My first introduction was at an event I attended on a whim because a doctor (MD) would be attending as a speaker. The talk revealed general adversarial networks (GANs) and their mind-boggling power and varieties, as well as, convolutional neural networks (CNNs) that could perform at the same discerning level as the average radiologist. It was incredible to think that with the laptop in my backpack I could design a neural network capable of discerning images at approximately the same level as a human who spent years honing their craft.

I took a 15 hour crash course I found on YouTube and learned all the fundamentals of how a neural network was built (its architecture) and how it “learned.” It was honestly quite a spectacle and rigorous. For anyone who wishes to enter the field, it is definitely necessary to at least have a foundational understanding of matrix algebra. Though understanding matrices and their operations are not necessary to create a model, it does facilitate the assimilation of new concepts and makes the whole idea of machine learning seem less whimsical and more mathematical.

In the end, the course only taught me the theory. It explained the power of CNNs and the mathematical operation called a “convolution” that they utilized. Though the course was definitely informative, it made me realize that the best way to learn is by “getting your hands dirty.” Working on a project forces you to actually understand the concepts and motivates you to learn. The field of machine learning is extremely vast. After getting the basics down of what constitutes a neural network, which only takes a few hours. Build. Experiment. Learn.

One of my favorite courses that adheres to this concept is FastAI which is a high-level package that is used to bolster the effectiveness and usage of the PyTorch library. It is too early to say anything, as I have only completed the first two chapters of the book and first three lessons of several. Nonetheless, it is an impactful, comprehensive course that truly motivates you with constant examples of application through Jupiter notebooks.

Time and deliberation will eventually make me capable of harnessing the utility of machine learning toward several projects in my undergraduate and future career.