Thank you for this series. Can you please share the equation for delta w and b updation each training example, same as previous question in the comment ?
@salahaldeen1751
Жыл бұрын
Thank you so much ! may I know how the delta W & delta b are computed each time, please?
@parimalarajamohan5370
2 жыл бұрын
Nice lecture I want to know whether the parameter may be partially removed or added in next batch (20%)
@SebastianRaschka
2 жыл бұрын
Thanks! Do you mean the model parameters? They don't get removed or added. Each batch updates them based on the loss gradient with respect to that parameter.
@rashutyagi9162
3 жыл бұрын
Nice Lecture Sir. I wanted to ask that if some algorithm makes more updates then it will be faster or slower? As minibatch makes more updates than normal full batch mode hence minibatch is slower or faster ?
@SebastianRaschka
3 жыл бұрын
There is a sweet spot. If you make the minibatch size too small, then it may increase the number of iterations to reach before convergence. Full batch mode is also not efficient because you only make one update per epoch. In general, minibatch mode a) has faster iterations and b) is faster for training but c) takes longer to process one epoch than full batch mode.
@rashutyagi9162
3 жыл бұрын
@@SebastianRaschka Thank you sir for answering. Understood the point.
@alexvass
Жыл бұрын
Thanks
@SebastianRaschka
Жыл бұрын
Thanks a lot @alexvass ! Glad to hear you got something useful out of this video!
Пікірлер: 10