search

jinxixiang / low_rank_wsi

Exploring Low-Rank Property in Multiple Instance Learning for Whole Slide Image Classification
10 stars 0 forks source link

NLP (Non-Local Pooling) #2

Open wuxchcandoit opened 10 months ago

wuxchcandoit commented 10 months ago

Hi! Thanks for your great work! I wonder why there is a learnable variable named S in class NLP, namely self.S= nn.Parameter(torch.Tensor(1, num_seeds, dim)) because the paper mentions that a bag feature is obtained through max pooling over the output of ILRA layers. But in class NLP, it seems that the bag feature is learnable.

jinxixiang commented 10 months ago

thank you for your interest! To get the global feature for classification, we found both max-pooling and learning pooling are applicable. However learnable pooling is more stable across different datasets. Thus, this implementation is slightly different from the paper.