Machine Learning Engineering Basics Tutorial

This material mostly in the form of Jupyter notebooks, is aimed at preparing undergraduate and graduate students for the very basics of practical machine learning. It starts with very simple Python exercises, moves on to Numpy and finally on to simple regression and classification toy problems using Tensorflow. The only prerequisites are basic programming experience and the basic concepts behind machine learning problems.

This is a work-in-progress and will be further improved in the coming semesters.

For any questions please contact [email protected]

Week 1

Content

• Python Basics (calculations, variables, functions, conditionals, loops and packages)
• Numpy (arrays, properties, calculations and operations)
• Tensorflow (basics, linear regression)

Goals

• Familiarize yourself with the basics of Python, Numpy and Tensorflow.
• Be able to build simple regression models with Tensorflow.
• Build a Sequential multi-layered model with Tensorflow and optimize it.

Sources

• [1] The Python Tutorial
• [2] Numpy Quickstart
• [3] Tensorflow and Keras Basic Regression
• [4] A line-by-line layman’s guide to Linear Regression using TensorFlow (adapted to TF2)

Week 2

Content / Goals

• Build simple regression models with Tensorflow.
• Build a simple machine learning pipeline (preprocessing, learning, evaluation)
• Build a classification multi-layer perceptron with Tensorflow.
• Extend it to use convolutional layers and observe the difference.

Sources

• [1] Tensorflow and Keras Basic Regression
• [2] A line-by-line layman’s guide to Linear Regression using TensorFlow (adapted to TF2)
• [3] Tensorflow Documentation: Loading MNIST
• [4] Basic classification: Classify images of clothing
• [5] Simple MNIST Convnet