Spiking Deep Residual Networks

This repo holds the codes for Spiking Deep Residual Networks.

Dependencies

• MATLAB
• MatConvNet

How to use

• Installing and compiling MatConvNet
• Merge the matconvnet directory with installed MatConvNet library.

MNISTS

• Train ANN models
  run matconvnet/examples/mnist/mytrans_mnist_beta.m
• Find activations for weight normalization.
  run matconvnet/examples/mnist/find_activation_single_gpu.m
• Conversion and test.
  run matconvnet/examples/mnist/ann2snn.m

CIFAR-10

• Train ANN models
  run matconvnet/examples/cifar10/mytrans_mnist_beta.m
• Find activations for weight normalization.
  run matconvnet/examples/cifar10/find_activation_single_gpu.m
• Conversion and test.
  run matconvnet/examples/cifar10/ann2snn.m

CIFAR-100

• Train ANN models
  run matconvnet/examples/cifar100/mytrans_mnist_beta.m
• Find activations for weight normalization.
  run matconvnet/examples/cifar100/find_activation_single_gpu.m
• Conversion and test.
  run matconvnet/examples/cifar100/ann2snn.m

ImageNet

• Use pre-trained ANN models from this link.
• Find activations for weight normalization.
  run matconvnet/examples/imagenet/find_activations_single_pt.m
• Conversion and test.
  run matconvnet/examples/imagenet ann2snn_res18_centre_pt.m