PocketFlow

PocketFlow copied to clipboard

• Doc Link: https://pocketflow.github.io/cpr_learner/

Describe the documentation issue There is a little problem in the algorithm description section. When it comes to "The matrix multiplication can be decomposed along the dimension of input channels. ", the size of sub-matrices {W_i} should probably be h_kw_k × c_o, not h_kw_kc_i× c_o, isn't it?

Thanks for pointing this out. You're right. We will change from

We divide $\mathbf{X}$ into $c_{i}$ sub-matrices $\left{ \mathbf{X}{i} \right}$, each of size $N h{o} w_{o} \times h_{k} w_{k}$, and similarly divide $\mathbf{W}$ into $c_{i}$ sub-matrices $\left{ \mathbf{W}{i} \right}$, each of size $h{k} w_{k} c_{i} \times c_{o}$.

to

We divide $\mathbf{X}$ into $c_{i}$ sub-matrices $\left{ \mathbf{X}{i} \right}$, each of size $N h{o} w_{o} \times h_{k} w_{k}$, and similarly divide $\mathbf{W}$ into $c_{i}$ sub-matrices $\left{ \mathbf{W}{i} \right}$, each of size $h{k} w_{k} \times c_{o}$.