NeuralNetw​orkClassif​ication object not supported for code generation

I trained a model using fitcnet() as follows:

mdl = fitcnet(trainTable,"Labels","Standardize",true,"Verbose",1)

I would like to see what the C/C++ code looks like so I tried

saveLearnerForCoder(mdl,'fitcnetCode')

but got the following error message:

First argument must be a classification, regression, or nearest neighbor searcher model supported for code generation.

Looking at https://www.mathworks.com/help/stats/savelearnerforcoder.html#bvclu99-Mdl models generated by fitcnet should work as they are ClassificationNeuralNetwork objects. Seems like a straightforward implementation. Why do I get that error?

Answer 1

[Ji Lee]

MATLAB Coder code generation support for neural networks became available recently in R2022a:

https://www.mathworks.com/help/stats/release-notes.html?category=code_generation&rntext=&startrelease=R2021b&endrelease=R2022a&groupby=release&sortby=descending&searchHighlight=

That particular documentation link points to the documentation for the latest release (R2022a). If you are using R2021b, you'll want to examine the archived R2021b version of that documentation:

https://www.mathworks.com/help/releases/R2021b/stats/savelearnerforcoder.html?s_tid=doc_ta#bvclu99-2