This repository contains presentation material for WDSS's 'Opening the Black Box' talk, held virtually on the 10th June 2020. The original stream can be re-watched here. The talk description is as follows:
After the 2008 financial crash, society has demanded more than ever that financial and tech institutions operate with transparency. How does this transfer to the age of AI and machine learning in which opaque, black box models rule? In this talk, three Warwick students offer a quick, high-level introduction to a handful of ways in which we can interpret such models, to allow us to perform data science ethically and openly.
The first speaker is James Davey, a first year maths student, who will discuss how we can peer into the 'brains' of computer vision systems.
The second speaker is Tim Hargreaves, a second year statistician, who will demonstrate how a technique from game theory can be used to answer questions about model decisions.
The final speaker is Farhan Tariq, a third year computer scientist, who will introduce a technique known as LIME for explaining model predictions.
You can download the presentation material above, or using the following links:
- James Davey — Interpreting Convolutional Neural Networks
- Tim Hargreaves — A Brief Introduction to Shapley Values
- Farhan Tariq — Interpreting with LIME
To keep up-to-date with WDSS, please follow us on our social media accounts: