There is one thing which I don't really understand:
if we use BN layer in recursive model, means we are going to use the same BN layer for every recursive block an unit. But conv-layer output at different depth is difffernt and thus need different parameter to do the BN...but we apply the same BN again and again...thus what does the BN layer actually do here?
Really thanks if you can give me some hints about this point.
Hi, @tyshiwo
Thanks for your papar and code, really like it.
There is one thing which I don't really understand: if we use BN layer in recursive model, means we are going to use the same BN layer for every recursive block an unit. But conv-layer output at different depth is difffernt and thus need different parameter to do the BN...but we apply the same BN again and again...thus what does the BN layer actually do here?
Really thanks if you can give me some hints about this point.