Nepal AI Winter School Lecture Notes

Last month, I prepared a series of lectures for the (really wonderfully organized) first Nepal AI Winter School.

The main talks were,

Deep Learning Foundations
Humanitarian AI
Useful Layers
Bayesian Deep Learning
GANs
Metalearning
Interpretability

The lecture videos are online here.

I also prepared a short exercise sheet (solutions).

In retrospect, I would have given a more visual introduction to VAEs. As a placeholder until I write about this in depth, here are some notes I kept from a discussion with students from a lab session during the school. I think the most useful pair of pictures was the per-pixel gaussians from the encoder (first page bottom left) and latent-space axis-aligned gaussians (second page top right)