Randperm and weight initialization of a neural network in parfor loop

1 次查看（过去 30 天）

显示更早的评论

Jack 2014-9-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/154380-randperm-and-weight-initialization-of-a-neural-network-in-parfor-loop

编辑： Jack 2014-9-11

在 MATLAB Online 中打开

Hi.

These is a discussion here. please read it :

http://stackoverflow.com/questions/25781336/using-rngshuffle-function-in-parallel-computing/25783711?noredirect=1#comment40326129_25783711

I have weight initialization with `configure` and `randperm` function in `parfor` loop. can i use only `randperm` and `configure` or I should use something likes:

    s = RandStream.create('mrg32k3a','NumStreams',4,'Seed','shuffle','CellOutput',true)

before `parfor` loop? .I'm calling my `parfor` loop again and again from another m-file. I want different random numbers and weights in every calling.

Thanks.