hassony2
commented
4 years ago Hi !
We did not experiment with supervising theta in this case, which could be or not beneficial, I would have to try ! :) Although the vertex prediction is done in a larger space, the network still outputs the low-dimensional PCA parameters, and then the higher dimensional vertices are obtained in a deterministic way. Therefore, it is unclear to me whether supervising the theta parameters should perform better :)
Best,
Yana
Thank you for great project. I read your paper recently. Why not supervised theta In Sec3.1 since the more supervised, the higher accuracy in appendix A1. And regress theta Easier than vetex because smaller space. So, I gauss expend theta will obtain a good result or not too bad.