lekhovitsky/aml-class-20

UCU AML 2020

Automated Machine Learning

This repository holds the slides and examples of a class on Automated Machine Learning as Jupyter notebooks.

Requirements

• Python 3.7 and pip (see guides on installation for your operating system)
• Jupyter
• Optional, but recommended: virtualenv
• Optional, but recommended: virtualenvwrapper
• Optional: Graphviz

Setup

Clone repository

git clone https://github.com/vikua/aml-class-20.git
cd aml-class-20

Install with virtualenv & virtualenvwrapper

mkvirtualenv aml

Optiopnal step, env should be activated by default once created:

workon aml

Install dependencies:

pip install -r requirements.txt

Create ipython kernel:

python -m ipykernel install --user --name aml

Install with virtualenv only

virtualenv aml
source aml/bin/activate
pip install -r requirements.txt
python -m ipykernel install --user --name aml

Simple install using default python & pip

pip install -r requirements.txt

Start Jupyter Notebook

1. At the command line, run jupyter notebook
2. Open your web browser to the directed url
3. Open ipynb file of interest

Table of content

1 Introduction

1.1 What is Automated Machine Learning?

1.2 Machine Learning 101

1.3 Model Selection and Assessment

1.4 Hyper-parameter Tuning

1.5 Debugging and Improving ML Models

II Tools & Techniques for Automated Machine Learning

2.1 Machine Learning Pipelines

2.2 Bayesian Hyper-parameter Optimization

2.3 Pipeline Optimization

2.4 Advanced Topics

III Create Value using Automation

3.1 Democratizing Machine Learning

3.2 Model Factories

3.3 Continous Learning