Full session (30 minutes)
Deep learning
Machine Learning

With deep learning and computer vision, there are many pitfalls and hacks to work around and debug. On June 30th, 2018, Andrej Karpathy, Director of AI at Tesla, tweeted a short list of first things to check when your neural network isn’t working.

In this session, you will see what these mistakes look like in code and performance metrics. Using a computer vision dataset and a PyTorch code sample - we’ll walk through each of these pieces of advice, test it and explain it. Expect a technical deep dive and a review of best practices when debugging a PyTorch computer vision experiment.

Original tweet at: https://twitter.com/karpathy/status/1013244313327681536

Yuval Greenfield