AI Music Generation with LSTM

Project Overview

This system trains on classical MIDI files to learn musical patterns, then generates new compositions using a stacked LSTM neural network architecture.

Model Architecture

3 LSTM layers (512 units each)
Dropout (0.3) for regularization
Dense layers with softmax output
Trained for 20 epochs with Adam optimizer