cnn

CNN model in pytorch .. rubric:: References

[1] Reddi S, Charles Z, Zaheer M, et al. Adaptive Federated Optimization. ICML 2020. https://arxiv.org/pdf/2003.00295.pdf

Module Contents

CNN_FEMNIST	Used for EMNIST experiments in references[1]
CNN_MNIST
CNN_CIFAR10	from torch tutorial
AlexNet_CIFAR10

class CNN_FEMNIST(only_digits=False)

Bases: torch.nn.Module

Used for EMNIST experiments in references[1] :param only_digits: If True, uses a final layer with 10 outputs, for use with the

digits only MNIST dataset (http://yann.lecun.com/exdb/mnist/). If selfalse, uses 62 outputs for selfederated Extended MNIST (selfEMNIST) EMNIST: Extending MNIST to handwritten letters: https://arxiv.org/abs/1702.05373 Defaluts to True

Returns:

A torch.nn.Module.

forward(x)

class CNN_MNIST

Bases: torch.nn.Module

forward(x)

class CNN_CIFAR10

Bases: torch.nn.Module

from torch tutorial https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html

forward(x)

class AlexNet_CIFAR10(num_classes=10)

Bases: torch.nn.Module

forward(x)