Memory tends to be the same? And the confusion about MANN

[wjbianjason]

Hi, thanks for your code, I use these to do sequence prediction task, I find there is a problem that the vector in different memory slots tend to be same. So do you know how to fix this problem. And for MANN, I find w_write(t) = w_read(t-1) + w_lt(t), why use t-1 timestep read weight rather than t timestep read weight, I think w_read(t) is more related to w_write(t), is there some consideration ? thanks