I had a confusion about SGD and many resources on the net added to that confusion. It was about SGD. From the view point of statistic, the term stochastic is used to indicate a random sample out of multiple samples. So one can easily confuse that SGD is faster because it randomly picks one sample out of a batch. While this is still correct for SGD, in practice SGD applies gradients immediately after each sample is processed. The reason for this is that SGD treats each sample as a batch. This was cleared to me by Jason Brownlee when I asked a question to him. Many thanks to Jason!
Subscribe to:
Post Comments (Atom)
-
Most of the google tutorials on keras do not show how to display a confusion matrix for the solution. A confusion matrix can throw a clear l...
-
This error means you have created the DbContext but not configured/added it to the project using either DI in startup.cs or by using DbCon...
-
This happens when you dont define primary key for an entity type. Define a primary key either explicitly or implicitly.
No comments:
Post a Comment