PyTorch 實現kmax-pooling

max-pooling有很多種實現方式

（1）kmax-pooling

import torch
def kmax_pooling(x, dim, k):
    index = x.topk(k, dim=dim)[1].sort(dim=dim)[0]
    return x.gather(dim, index)
x = torch.rand(4, 5, 6, 10)
print(x)                        # [torch.FloatTensor of size 4x5x6x10]
y = kmax_pooling(x, 3, 5)       
print(y)                        # # [torch.FloatTensor of size 4x5x6x5]