Skip to content

Learning a hierarchical model of data with neural networks.

License

Notifications You must be signed in to change notification settings

pcsl-epfl/hierarchy-learning

Repository files navigation

Hierarchy Learning

Implement the Random Hierarchy Model (RHM) in PyTorch and train (deep) neural networks or kernels on it.

Code for the paper "How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model", preprint available here: arXiv:2307.02129.