- Example 1: MNIST classification with PNNs In this notebook, we give an introduction to physical neural networks (PNNs) by showing how one can construct a two-layer PNN with coupled nonlinear oscillators to perform the MNIST task.
- Example 2 Fashion MNIST classification with PNN and PAT In this notebook, we show how a physical neural network can be constructed out of coupled nonlinear oscillators to perform the Fashion MNIST task. We introduce a simulation-reality gap and show how to train the PNN with PAT.
- Example 3: Toy example of a PAT autodifferentiable function This notebook contains a simple implementation of PAT and demonstrates some of PAT's advantages over in silico training on a toy example. It is most interesting for users who want to see an implementation of PAT with as little coding overhead as possible.
