sparse autoencoder

a sparse autoencoder1 is running live in your browser — model trained on mnist with top-20 sparsity via tf.js. inspired by this and that.

hidden layer units grid2
top-4 activating inputs3
feature —
  1. Sparse Autoencoders (SAEs) are typical autoencoders that enforce a sparsity constraint on their hidden units, which are typically of higher dimensionality than their input. This is motivated by the superposition hypothesis: neurons hold information about multiple concepts that are rarely simultaneously present. SAEs are frequently used for mechanistic interpretability today.
  2. This 40x40 grid gives the average activation value of the 1600 hidden units in this SAE, computed across a sample of the entire MNIST dataset. Every two seconds, a new hidden unit is chosen.
  3. These four images displayed are the images which activated the current hidden layer activation unit the most. It might be fun to see if there is a common pattern among these images.