Partition

[LeiJiang1]

Hello, I can only see that you use joint_label as partition matrix while you are supposed to use partition strategy to learn the partition matrix. Do I miss anything? If so, could you let me where your learnable partition matrix code is. Thanks!

[ZhouYuxuanYX]

Hi,

Thank you for your interest in our work! As described in Sec. 4.4 of our paper, learning of the partitions is easily achieved by predefining a parameterized partition matrix and apply softmax on it to represent the "soft" partitions. Just replace the "label" variable in line 423 in model/Hyperformer.py with the parametrized "soft" partition matrix and train the model until converge. The learned partition matrix can be binarized and used as label in the normal training setup. We omit this part of code since we already provide the learned labels.