ERROR: A regression layer must not be preceded by a softmax layer.

16 views (last 30 days)
hello, I want to ask why a regression layer must not be preceded by a softmax layer? If I want to add a softmaxlayer before regressionlayer, what should I do?

Answers (1)

Greg Heath
Greg Heath on 13 Apr 2018
Softmax is designed to be a classifier output component that yields a classification probability output with components in [0 1 ] and column sums of 1.
Combining the output of multiple classifiers is the only configuration I can think of where the softmax layer may not be the output layer.
Hope this helps.
Thank you for formally accepting my answer
Greg
  1 Comment
Rotem Mairon
Rotem Mairon on 7 Oct 2019
This is true, I think. However, if you try to run a network with the last layer being softmax, in Matlab you would get the following error message: "Missing output layer. The network must have one output layer." So how can one use softmax as output?
Thanks for your time.

Sign in to comment.

Categories

Find more on Deep Learning Toolbox in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!