Neural Tunes: AI Music Generator 🎵

Overview

Neural Tunes is an advanced machine learning project that generates original music sequences using Long Short-Term Memory (LSTM) neural networks. The system learns patterns from MIDI files to create new musical compositions, demonstrating the potential of AI in creative applications.

Performance Metrics

Training Accuracy: ~10%
Validation Loss: 5.67
Total Parameters: 822,400
Training Data: 14,495 notes processed
Unique Notes: 174

Key Features

• Preprocessing Pipeline
  MIDI file parsing using Music21
  Note sequence generation
  One-hot encoding for musical notes
  Data normalization
  
  

• Model Architecture
  Sequential note prediction
  Temperature-based sampling
  Dropout layers for regularization
  Adam optimizer with categorical cross-entropy loss
  
  

• Generation Capabilities
  Multiple music styles (balanced, creative, structured)
  Variable sequence lengths
  Adjustable creativity parameters