neuroptica

neuroptica copied to clipboard

Added trainable nonlinearity class with example

2

Hey Ben, I got in contact with you about a year ago about using this framework, and I ended up using it in my masters thesis. I extended it to...

bendavidsteel

This introduces automatic computation of the backpropagation through activation functions. Changes: There is only one `Nonlinearity` class. Real, complex, and non-holomorphic activations handled exactly the same way. You only need...

twhughes

loss function nan when using AbsSquared activation

1

running this mode ``` model_linear = neu.Sequential([ neu.ClementsLayer(N), neu.Activation(neu.AbsSquared(N)), neu.DropMask(N, keep_ports=range(N_classes)) ]) losses = neu.InSituAdam(model_linear, neu.CategoricalCrossEntropy, step_size=step_size).fit(x_train_flattened, y_train_onehot, epochs=n_epochs, batch_size=batch_size) ``` gives the warning: ```../neuroptica/neuroptica/losses.py:45: RuntimeWarning: overflow encountered in exp...

twhughes

IndexError when trying to train a network with N = 2

1

Not sure if I am doing something stupid, but I get the following error when trying to train a mesh of dimension N = 2. ``` --------------------------------------------------------------------------- IndexError Traceback (most...

ianwilliamson