Neural Network Library (Obsolete)

Resources Community Resources
,
#227

My tiny brain cannot comprehend yet :flushed:.

1 Like
#228

Could it be because the reward being below 0? I have experienced this couple of times too but due to how KNN calculates stuff based of scores, it always have to be above 0. try clamping the wins, or perhaps lower the rate of giving punishments

1 Like
#229

i fixed the that not popping up, but now it just has nil for this


	local min,max = scores[best[#best]],scores[best[1]]
	local range = max-min
#230

I know why the nets[best[1]] is nil
because nets is the GenNet Folder and it has 1 array only
if the value is higher than 1 then it will return to nil

i tried print(nets[1])
and it returns ā€œGenNet_1ā€

#231

ā€¦this is exactly why I am retiring this module. I made it before I discovered dictionaries and my use of arrays here is absolutely disgusting. The user shouldnā€™t have to guess what they are getting when they put a bunch of numbers into a 4 dimensional array.

5 Likes
#232

Quick update on the new library! Iā€™ve finally gotten past the main roadblock of the development (LSTM backpropagation) and am speeding through the remaining portions of the libraryā€™s features.
Currently, this is what the package structure looks like:

fesad
fesad285Ɨ506 8.5 KB

Yes, genetic learning is next on the list.
Iā€™ll soon post a thread where I want any NN enthusiasts to post any features they would like to see in the module as I do not want anyone to be left out. I will link the thread here when I make it.

EDIT: I posted the thread here.

Stay safe out there.

4 Likes
#233

Iā€™m hype for the new library so I can continue my project

4 Likes
#234

Iā€™m also hyped as Iā€™m excited about this module.

2 Likes
#235

Ik Youā€™re Working on a new module, But I have found out how to make a successful NN without it errorā€™ing. Iā€™ll send the place file. It works like Jabrils NN where the Ai controls a character thatā€™s moving forward based on its direction. (I was gonna use coroutine but I donā€™t know a way to kill them) I also couldā€™ve used better code. nn.rbxl (45.9 KB)

2 Likes
#236

Wow this is amazing! definitely will be useful, thank you for this resource!

1 Like
#237

Forgot to write here that the second library was released a few days ago. Doesnā€™t seem like people are nearly as interested anymore but if you want to check it out, go ahead.

5 Likes