camille-vanhoffelen / modern-ml-engineer

Introduction to Machine Learning - Ilia State University

🎓 The Modern Machine Learning Engineer

🤗 Welcome!

This is a course which teaches you how to build useful ML software.

This course aims wide. We will cover all topics required to be a successful ML engineer. This includes data engineering, data exploration, data analysis, and ML engineering, and frameworks like numpy, pandas, sklearn, keras, and pytorch (for a full list check the syllabus and the learning goals section).

This course is lean. You will learn just enough to build AI solutions from scratch. ML is a vast subject however, and resources are provided for deeper dives into any of its fields.

This course is pragmatic. All lectures consist of slides explaining key theoretical concepts, followed by a hands-on python notebook with coding exercises.

✅ What this course will do:

• teach you all the theory and skills required to create effective, scalable, robust, ethical machine learning systems
• give you hands-on experience training, improving, and deploying ML models
• make you employable in the AI industry
• turn you into a jack of all trades
• guide you to become a master of few
• share some low quality AI memes

❌ What this course won't do:

• give you 10 years of experience in ML
• turn you into a Deep Learning wizard
• make you publish a paper at NeurIPS
• build Skynet

🤔 Why this course?

What once was a niche research topic has blossomed into a mature engineering field. There are excellent online degrees that focus on ML theory, and great practical courses that cover frameworks. But the barrier remains high to start a career in AI, as none of these teach you how to build ML products from scratch. This course's purpose is to enable all coders to get out there and solve the world's problems one neural network at a time.

👽 Who is this for?

This course is perfect for the software engineer curious about ML technologies, the data scientist looking to move past Kaggle kernels, or the AI enthusiast wondering how systems are built in the real world.

Beginner programming skills are required. A little python experience (can you define a function?), and some statistical basics (what's a standard deviation?) are recommended.

👨‍🏫 Who is this by?

Camille Van Hoffelen has worked as an ML engineer for the past 6 years, with a focus on large scale Natural Language Processing systems. This course was taught at Ilia State University, Georgia, in 2020.

🚀 How should I use this course?

Here's a few ideas:

• For each lecture, read the slides, then go through the notebooks. Complete the 💪 and 🧠 exercises, then flip through the additional resources.
• Skip straight to a particular section/lecture if you have already taken 50 billion ML courses.
• Forget about the slides, find the notebook for that method you can't remember how to use, and copy paste to your heart's content.
• Test yourself with the assignments. Become the nerdiest Pokemon trainer there ever was.

Notebooks can be viewed in github, viewed in nbviewer, run locally with jupyter, run with mybinder, or run with google colab.

Syllabus

Introduction

• Workstation Setup
  slides notebook

Chapter 1: Data Engineering

• 1.1 Fundamentals of Data Engineering & Data Lakes
  slides notebook
• 1.2 Data Pipelines
  slides notebook
• 1.3 Data Warehouses
  slides notebook
• 1.4 Effective Data Storage
  slides notebook

Chapter 2: Data Exploration

• 2.1 Introduction to Numpy & Pandas
  slides notebook
• 2.2 Tabular Data Pt.1
  notebook
• 2.3 Tabular Data Pt.2
  notebook
• 2.4 Time Series Data
  slides notebook
• 2.5 Text & Image Data
  notebook
• 2.6 Data Visualization
  slides notebook

Chapter 3: Data Analysis

• 3.1 Introduction to Machine Learning & Clustering
  slides notebook
• 3.2 Dimensionality Reduction
  slides notebook
• 3.3 Anomaly Detection
  slides notebook
• 3.4 Supervised Learning Fundamentals
  slides notebook
• 3.5 Linear Regression
  slides notebook
• 3.6 Logistic Regression
  slides notebook
• 3.7 Learning Better Pt.1
  slides notebook
• 3.8 Learning Better Pt.2
  slides notebook
• 3.9 Support Vector Machines
  slides notebook
• 3.10 Random Forests
  slides notebook
• 3.11 Neural Networks Pt.1
  slides notebook
• 3.12 Neural Networks Pt.2
  slides notebook
• 3.13 Learning Better Pt.3
  slides notebook
• 3.14 Introduction to Pytorch
  slides notebook
• 3.15 Computer Vision Pt.1
  slides notebook
• 3.16 Computer Vision Pt.2
  slides notebook
• 3.17 Natural Language Processing Pt.1
• 3.18 Natural Language Processing Pt.2
• 3.19 ML Research

Chapter 4: ML Engineering

• 4.1 Fundamentals of ML Engineering
  slides notebook
• 4.2 Evaluation Pt.1
  slides notebook
• 4.3 Evaluation Pt.2
  slides notebook
• 4.4 Error Analysis
• 4.5 Hyperparameter Optimization
• 4.6 ML Challenges
• 4.7 ML Labs
• 4.8 Datasets & Labeling
• 4.9 ML Architecture
• 4.10 ML Deployment
• 4.11 Data Privacy & Security
• 4.12 ML Ethics

Assignments

• Assignment 1
• Assignment 2
• Assignment 3
• Assignment 4
• Final Project

Learning Goals

This course is designed to step through the journey of building a production data application, as reflected in the following learning goals:

• Storing and Managing Data
  Students can build data lakes and data repositories, as well as data pipelines to automate the gathering and processing of data.
• Exploring and Analysing Data
  Students can describe, visualise, and provide insights into tabular, time series, image, text, and geospatial data.
• Building a Machine Learning Product
  Students can build useful classification, regression, and clustering tools using Machine Learning methods.
• Being Ethical and Secure
  Students understand the importance of privacy in data driven methods, and the responsibility of engineers concerning the impact of these technologies.

Assignments

The coursework is split between four small assignments, a final project, and a final presentation.

The small assignments serve to synthesise the previous course content on your own, and put it to practice. They are all coding exercises: you will be given resources and/or code stubs, and will submit runnable code and some observations (see assignments readme).

The final project tests everything that you have learnt from this course. This is a python notebook report like those data scientists make to share their experimental progress. It tests your ability to design, carry out, and communicate machine learning experiments. This is complemented by the final presentation, a 15mn talk to synthesize, and discuss the results. (see final project readme).

License

This work is licensed under a Creative Commons Attribution 4.0 International License. See the LICENSE.txt file for details.