E. Almqvist
|
ed0751a110
|
Tweaked inputs for training
|
4 years ago |
E. Almqvist
|
56c765d8bb
|
Huge tweaks to the network
|
4 years ago |
E. Almqvist
|
eab52714a4
|
Tweaked AI a lot, made tests in the end of training
|
4 years ago |
E. Almqvist
|
bfa7c451c7
|
Tweaked the network etc
|
4 years ago |
E. Almqvist
|
21f531669d
|
More learning rate stuff
|
4 years ago |
E. Almqvist
|
c0a536c49a
|
Optimzed stuff
|
4 years ago |
E. Almqvist
|
4a4b3b26bb
|
Added lr optimization inspired by the hessian matrix
|
4 years ago |
E. Almqvist
|
4f15e9e34c
|
Added .pyc to gitignore
|
4 years ago |
E. Almqvist
|
e05ad561ea
|
Added experimental lr function and more bugs
|
4 years ago |
E. Almqvist
|
bd7c52c927
|
Feed the AI random inputs per learn iteration
|
4 years ago |
E. Almqvist
|
eac56f983d
|
Made the AILib into a module
|
4 years ago |
E. Almqvist
|
606e473d1e
|
Added more bugs and lr is now relative to the gradient
|
4 years ago |
E. Almqvist
|
2925776052
|
Small tweaks
|
4 years ago |
E. Almqvist
|
2f1d7d458d
|
Tweaked the cost function
|
4 years ago |
E. Almqvist
|
6d5fc63e05
|
Made the gradient calculations much better
|
4 years ago |
E. Almqvist
|
0a754de2e9
|
Added instance compairing for bias and weights
|
4 years ago |
E. Almqvist
|
3c9bffd797
|
Stuff
|
4 years ago |
E. Almqvist
|
2a597c1a34
|
Optimized code and added more bugs
|
4 years ago |
E. Almqvist
|
0c6d554268
|
Cleaned code
|
4 years ago |
E. Almqvist
|
ab6d2a1169
|
Gradient stuff
|
4 years ago |
E. Almqvist
|
ff3d69bf61
|
Switched to tabs
|
4 years ago |
E. Almqvist
|
a543a9db1a
|
Made it so that the AI trains on the same input
|
4 years ago |
E. Almqvist
|
c1de21eaad
|
Stuff
|
4 years ago |
E. Almqvist
|
bce25a5a44
|
It works
|
4 years ago |
E. Almqvist
|
bff87e11a4
|
Fixed small bug
|
4 years ago |
E. Almqvist
|
bc2e219108
|
Added ai that works i think
|
4 years ago |
E. Almqvist
|
2dec1df539
|
Added debug because I have no clue if it is working
|
4 years ago |
E. Almqvist
|
fc779f1afd
|
Backprop and gradient stuff
|
4 years ago |
E. Almqvist
|
788ee4fb2d
|
Made so that the program runs
|
4 years ago |
E. Almqvist
|
1493824ac7
|
Removed small bugs
|
4 years ago |
E. Almqvist
|
59fc19e181
|
Added more bugs
|
4 years ago |
E. Almqvist
|
d5a181ba39
|
Bug fix
|
4 years ago |
E. Almqvist
|
9cb11da281
|
Fixed bugs and fixed cost func
|
4 years ago |
E. Almqvist
|
bb7297a676
|
Added recursive backprop
|
4 years ago |
E. Almqvist
|
d928c454e1
|
Small fix
|
4 years ago |
E. Almqvist
|
5a88f515d2
|
Gradient stuff
|
4 years ago |
E. Almqvist
|
7bee53c1f8
|
Gradient and learning progress
|
4 years ago |
E. Almqvist
|
df55c8eaa7
|
Stuff
|
4 years ago |
E. Almqvist
|
3587e223d5
|
Removed stuff
|
4 years ago |
E. Almqvist
|
4905f4210f
|
Tweaked thinking
|
4 years ago |
E. Almqvist
|
90071935a7
|
Changed param datatype
|
4 years ago |
E. Almqvist
|
012b3a12a1
|
Stuff
|
4 years ago |
E. Almqvist
|
52054ad667
|
Small changes
|
4 years ago |
E. Almqvist
|
d4475cd4a0
|
Borked gradient calculations
|
4 years ago |
E. Almqvist
|
3081b06253
|
Stuff
|
4 years ago |
E. Almqvist
|
26d8e06230
|
Removed unneeded code
|
4 years ago |
E. Almqvist
|
6ebaf94869
|
Stuff
|
4 years ago |
E. Almqvist
|
544f6e454c
|
Learning that doesn't work
|
4 years ago |
E. Almqvist
|
86727591df
|
Small fix
|
4 years ago |
E. Almqvist
|
80013cab71
|
Gradient calculation
|
4 years ago |