Clone this repo:

Branches

  1. 00b2668 final gds oasis by Jeff DiCorpo · 6 weeks ago main
  2. 31e7298 Add timing information by Tamas Hubai · 7 weeks ago
  3. 6d55480 Harden user_project_wrapper by Tamas Hubai · 7 weeks ago
  4. 4e9bdf5 Manually edit obstructions in LEF by Tamas Hubai · 7 weeks ago
  5. 6926092 Harden trainable_nn by Tamas Hubai · 7 weeks ago

Trainable Neural Network

Overview diagram

Implements a simple neural network that supports on-chip training in addition to inference. The two hidden layers use leaky ReLU as their activation function while the output layer uses a rough approximation of softmax.

Unlike usual neural network implementations we use fixed-point saturation arithmetic and each neuron and synapse is represented as an individual instance. Inputs, outputs and weights, as well as forward and backward propagation can be managed through the wishbone bus.