Hi,

This week you'll learn about Automatic Differentiation Part 2: Implementation Using Micrograd.

Image

Every person in Deep Learning has used either PyTorch, TensorFlow, or a different Deep Learning framework. So why not build your own?

Andrej Karpathy (formerly the director of Tesla AI and a lifelong Machine Learning Educator) proposed this question in one of his YouTube videos. The result? A Deep Learning framework built from scratch that can calculate gradients automatically, thereby being able to train an Artificial Neural Network (from scratch). He calls it micrograd.

The big picture: micrograd is a bare-bones Python package capable of automatic gradient calculation. It works on scalar values and is hypothesized (by Andrej) to be the only thing one will ever need to train a neural network (everything else is just a scaling-up mechanism).

How it works: In this tutorial, we cover the ins and outs of the Python package. We also add our analysis and observations about the package.

Our thoughts: micrograd is a minimal Python package built for educational purposes (as Andrej writes in the README of the project). We wanted to build a tutorial around it for one to revisit Neural Network training from scratch. Knowing what happens under the hood of a Deep Learning framework will always make you go that extra mile.

Yes, but: Understanding the basics can be very theory intensive. With this tutorial, we wanted to build the intuitions with a code-first approach.

Stay smart: Do not learn concepts like object detection or classification at the surface level. Instead, go deeper and be future-proof with your investment in learning the fundamentals.

Click here to read the full tutorial

Do You Have an OpenCV Project in Mind?

You can instantly access all of the code for Automatic Differentiation Part 2: Implementation Using Micrograd, along with courses on TensorFlow, PyTorch, Keras, and OpenCV by joining PyImageSearch University. 

Guaranteed Results: If you haven't accomplished your Computer Vision/Deep Learning goals, let us know within 30 days of purchase and get a full refund.

Do You Have an OpenCV Project in Mind?



Your PyImageSearch Team