Questions about disentangled representation learning

[Ray-Zhen]

Hello,
Thank you for the nice work.
I have a question on the representation projection. In your paper, the text and video representation are independently project into K components with trainable parameter W, but in your implemention, you just split the representation with dimension transformation like:

        t_feat = cls.view(a, self.config.center, -1)
        v_feat = video_feat.view(a, b, self.config.center, -1)

It would be greatly appreciated if you could provide a explanation.
Thanks in advance for your time and assistance.
Best regards