search

cmhungsteve / TA3N

[ICCV 2019 (Oral)] Temporal Attentive Alignment for Large-Scale Video Domain Adaptation (PyTorch)
https://arxiv.org/abs/1907.12743
MIT License
259 stars 41 forks source link

Some questions about the paper #8

Closed DanZhang123 closed 4 years ago

DanZhang123 commented 4 years ago

Why is formula 7 a minus sign instead of a plus sign? Why not add more weight to the ones that are hard to distinguish.

I am looking forward to your answer. Thank you very much!

cmhungsteve commented 4 years ago

After we assign the attention weights, we will perform DA on all the attended data. Since we want our DA to focus more on the ones easy to distinguish in terms of domains (i.e. contribute more on to the overall domain shift), we use the minus sign here.

Hope that answers your question