Hello,
I was wondering why we don't use a convex activation (whose compositions would also be convex) wouldn't we then have a unique minimizer and then boom profit....
Thanks for clarifying :)
This can help, https://stats.stackexchange.com/questions/106334/cost-function-of-neural-network-is-non-convex
Convexity and NN
Hello,
I was wondering why we don't use a convex activation (whose compositions would also be convex) wouldn't we then have a unique minimizer and then boom profit....
Thanks for clarifying :)
1
This can help, https://stats.stackexchange.com/questions/106334/cost-function-of-neural-network-is-non-convex
Add comment