Computational Interaction 2016

Machine learning in HCI

Purpose

The purpose of today is to get hands on experience in applying machine learning to HCI. The focus is on specific, practical examples rather than theory.

Outline

We will cover three topics.

• Each topic is a condensed, applied exploration of machine learning in an HCI topic.
• Each topic lasts two hours.
• Each two hour block has the format:
  • 20 minutes lecture introduction
  • 30 minutes "getting to grips" practical
  • 20 minutes follow up lecture
  • 50 minutes challenge practical

The notebooks for each of the three summer school topics are listed below:

1. Classifiying Audio Streams This topic explores supervised classification of audio data, and how to evaluate classifiers without deceiving yourself.

2. Unsupervised Image Learning This topic will explore how high-dimensional sensor input can be organised with unsupervised learning to build primitives for interaction.

3. Inferring Typing Behaviour This exercise looks at building probabilistic Bayesian models of typing behaviour, and inferring parameters using Markov Chain Monte Carlo.