About this Project

Preface

At the end of this tutorial, we will be using the MNIST handwritten digit dataset to task our Convolutional Neural Network (CNN) for digit classification.

We will be releasing modules in a weekly fashion. The first module is already published, which is Week 1: Linear Classifiers.

Our plan is to first introduce Linear Classifiers through manual learning follow by Perceptron Basics and Multi Layer Perceptrons. After the learning the basics, we will be moving onto creating a basic Neural Network for recognizing digits, then a Fully Connected Neural Network, ending with a Convolutional Neural Network.

We’ll do our best to explain these concepts in a digestible way, but this tutorial’s purpose is for you to get your hands dirty with your first machine learning models. If you’re interested in learning more about the math and intricusies of each model, I would highly recommend attending our weekly workshops, where they go into much more detail.

More information about the workshops can be found at our AI@UCI discord.

The bare minimum knowledge needed for this self-guided project is proficiency in Python and basic data structures.

Usage

This website will be your tutorial and main source of information. You can work through the tutorial at your own pace. We will make an announcement on the Discord server and update the “Updates” section of the homepage when a new module is added

Please don’t hesitate to reach out to any of the Project Coordinators on Discord, we are here to help!

Codes will be provided through Google Colab files. This way, you can spend less time setting up your enviornment and get straight to learning about CNNs!

Friendly Submission/Leaderboard

Currently under construction

We will have a submission page where you can submit your model and we will evaluate your model with a hidden test set.

We expect to open submissions at around week 8/9, after we get to the CNN module.

Only the highest accuracy model of each participant will be kept.

Submission cooldown is 24 hours.

If you have any problems or notice any issues/typos, please reach out to one of the Project Coordinators.

Acknowledgements:

AI@UCI

Project Co-Leads: Sri Gubbala, Jonathan Pan

Project Coordinators: Harsh Sharaff, Pranav Sethia