search

matthewcarbone / Bootcamp

A collection of tutorials and resources for data science and machine learning
BSD 3-Clause "New" or "Revised" License
3 stars 2 forks source link

[DRAFT] Neural networks from scratch #6

Open matthewcarbone opened 2 months ago

matthewcarbone commented 2 months ago

Neural networks from scratch

It's very difficult to understand what a neural network is without coding one. We're gonna do that here.

Learning objectives

Content to cover

Capstone

Students will create their own deep learning "library" from scratch (or at least, they will begin this process). This library can be a true Python library (a .py file, if students are comfortable), or it can be a collection of Python functions and classes. It's completely up to the student. Nevertheless, this deep learning library should implement a simple neural network model, optimizer and any additional "bells and whistles" the students want. Students should test this framework on one of the problems they've encountered before.

dbiersach commented 2 months ago

Do you have a representative example of what you would define as a student's "success" in this task? Is this script indicative of the level of complexity you would want from the students? SimpleNeuralNetwork.zip

matthewcarbone commented 2 months ago

@dbiersach yes, your script is exactly what I have in mind.

Btw this module isn't going to be for the August bootcamp, this will be for hopefully future bootcamps 👍. I've outlined, loosely, the topics to cover for the August bootcamp in #7.

Also, I should note that part of this is about compiling existing resources. We don't necessarily have to do it all again. Check out the Intro to Python directory to see an example of that.