Version 1.0.2
Release Date: 2025-12-30

Initial Release
Implemented lightweight learnable interpolation using a compact MLP
Framework-free neural network (no PyTorch / TensorFlow)
Automatic data normalization (max / min-max)
Built-in learning rate finder

Training metrics:
Loss
RMSE
MAPE
R² score

Export support:
Pickle (.pkl) for Python inference
JSON weights for C++ / ESP32 deployment
Embedded-friendly weight format (raw weights & biases)
Python-only training, no GPU required
Prediction support from both PKL and JSON models
CSV export of actual vs predicted values
Auto-generated training and evaluation plots
Designed for edge devices and microcontrollers

