nnmf_24a/MLP_equivalent/README.md

The last 1x1 layer of the NNMF network is a problem and had to be removed.

The structure is now:

Sequential(
  (0): ReLU()
  (1): Unfold(kernel_size=(5, 5), dilation=(1, 1), padding=(0, 0), stride=(1, 1))
  (2): Fold(output_size=torch.Size([24, 24]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (3): L1NormLayer()
  (4): Conv2d(75, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (5): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (6): ReLU()
  (7): Conv2d(32, 32, kernel_size=(1, 1), stride=(1, 1))
  (8): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (9): ReLU()
  (10): Unfold(kernel_size=(2, 2), dilation=(1, 1), padding=(0, 0), stride=(2, 2))
  (11): Fold(output_size=torch.Size([12, 12]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (12): L1NormLayer()
  (13): Conv2d(128, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (14): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (15): ReLU()
  (16): Unfold(kernel_size=(5, 5), dilation=(1, 1), padding=(0, 0), stride=(1, 1))
  (17): Fold(output_size=torch.Size([8, 8]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (18): L1NormLayer()
  (19): Conv2d(800, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (20): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (21): ReLU()
  (22): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))
  (23): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (24): ReLU()
  (25): Unfold(kernel_size=(2, 2), dilation=(1, 1), padding=(0, 0), stride=(2, 2))
  (26): Fold(output_size=torch.Size([4, 4]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (27): L1NormLayer()
  (28): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (29): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
  (30): ReLU()
  (31): Unfold(kernel_size=(4, 4), dilation=(1, 1), padding=(0, 0), stride=(1, 1))
  (32): Fold(output_size=torch.Size([1, 1]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (33): L1NormLayer()
  (34): Conv2d(1024, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (35): ReLU()
  (36): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
  (37): ReLU()
  (38): Unfold(kernel_size=(1, 1), dilation=(1, 1), padding=(0, 0), stride=(1, 1))
  (39): Fold(output_size=torch.Size([1, 1]), kernel_size=(1, 1), dilation=1, padding=0, stride=1)
  (40): L1NormLayer()
  (41): Conv2d(96, 10, kernel_size=(1, 1), stride=(1, 1), bias=False)
  (42): Softmax(dim=1)
  (43): Flatten(start_dim=1, end_dim=-1)
)