use logitcrossentropy to train

This commit is contained in:
2023-05-28 22:08:08 +07:00
parent 8fd0cc0fdd
commit f849aea2a6

View File

@@ -41,7 +41,7 @@ using .learn
which defined by neuron.tau_m formula in type.jl which defined by neuron.tau_m formula in type.jl
Change from version: 0.0.2 Change from version: 0.0.2
[DONE] new learning method - new learning method
- use Flux.logitcrossentropy for overall error - use Flux.logitcrossentropy for overall error
- remove ΔwRecChange that apply immediately during online learning - remove ΔwRecChange that apply immediately during online learning
- collect ΔwRecChange during online learning (0-784th) and merge with wRec at - collect ΔwRecChange during online learning (0-784th) and merge with wRec at