数学 Discord 服务器开设免费机器学习教程与工作坊，10 月 17 日开课

This post is to let people know about a workshop/tutorial on Machine Learning that will run on the (blue) math discord server: discord.gg/maths.

The place is nice, people are cool and helpful, and as you can see it is a mathematics server, so the learning environment is quite focused indeed. We hope that you'll like the experience. Also, this is completely free, just in case anyone is wondering.

Timeline and Curriculum Overview

There will be weekly lectures and/or lab-sessions. The workshop is starting on Saturday, 17th October. The first lecture + lab session will be held from 3 to 5 PM CEST (UTC + 2). There will be quite a bit of coding involved on the first day, and the instructors will try to review people's codes.

For a curriculum overview, here you go:

Part 1: Traditional Machine Learning

• generalities, vocabulary
• linear models, gradient descent
• supervised learning methods
• unsupervised learning methods
• best practices in ML

Part 2: Introduction to Deep Learning

• the multi-layer perceptron and backpropagation
• auto-encoders
• recurrent neural networks for sequential data
• convolutional neural networks
• attention mechanism and the transformer
• signal processing

Part 3: graph machine learning

• traditional methods and feature engineering on graphs
• shallow node embeddings
• knowledge graph embeddings
• graph neural networks
• subgraph sampling
• more to come as we progress

Prerequisites

We think the most important part for you would be knowing how to code in Python (and use some libraries that are very commonly used like numpy or matplotlib for maths). That said, knowing calculus I (possibly calculus II) would be great, and I'll also link a few resources for basic linear algebra shortly.

Prerequisite 1 : Basic knowledge of Python:

Tutorial and Book (not really recommended because it's very long, but just for those who prefer).

Note that knowing a bit of OOP in Python (at least, how to handle classes) would be very nice.

Prerequisite 2 : Some useful libraries

For numpy, here's the official guide, and a book, and some video tutorials (about 1h).

For matplotlib here are some tutorials: matplotlib, w3schools. Also, Seaborn (optional, but oh boy it does look a bit nicer).

Prerequisite 3 : Proficiency with Calculus I

Here are some Calculus I resources and Calculus II resources (latter is optional, but sequences can be useful).

Prerequisite 4 (Optional) : A bit of basic knowledge in linear algebra

TBD. Still discussing what would be strictly necessary, and we don't really want to go too deep in order to keep the prerequisites accessible.

For any questions/queries that you might have, please write in the comments. I'll try to get back to you as soon as possible. Also, it's best if you join the server and directly talk to the instructors there.