v1.2.00

[mochji]

NEW FEATURES

• Support for custom activation functions
• Support for different activation functions on different layers
• Ability to customize amount of nodes in each layer
• Support for exporting neural network data
• New back-propagation algorithm
• New back-propagation algorithm with momentum

IMPROVEMENTS

• Code clean up (code is a lot better now like i mean a lot better)
• Weight, bias and current tables are now numbered instead of something like w3 or c1
• Output tables are now numbered instead of ow or oc
• Functions are now grouped into tables inside of lnn

BUG FIXES

• lnn.initialize() didn't fill the current tables. (this was the biggest bug ever, it affected everything that involved the neural networks)
• probably more but i forgot

DOCUMENTATION

• Complete documentation overhaul
• Documentation updates to include new functions, remove removed functions and correctly describe changed functions.

---

This discussion was created from the release v1.2.00.