Music Sheet Generator

CS 4100 - Artificial Intelligence | Northeastern University

Instructor: Rajagopal Venkatesaramani

Team: Mehr Singh Anand (Project Manager), Mihalis Koutouvos, William Sartorio, Nalini Singh

---

Usage

Make sure you are in the root folder.

Install dependencies:

pip install -r requirements.txt
pip install streamlit

Run app:

streamlit run app.py

To test endpoint(s), go to http://127.0.0.1:8000/docs after running the app

Project Statement

Traditional chord progression follows patterns that simultaneously satisfy statistical regularities (e.g., transition probabilities between chords) and theoretical constraints (e.g., voice leading, resolution, tension). Existing work tends to optimize a single technique in isolation. This project directly compares two fundamentally different AI approaches applied to the same task, offering insight into the tradeoffs between learned pattern recognition and rule-guided optimization.

---

Approach

Markov Chain Model

Models chord transitions as conditional probabilities derived from observed sequences. Captures what composers statistically do based on training data.

Genetic Algorithm Model

Encodes chord progressions as candidate solutions and evolves them over generations using a fitness function grounded in music theory rules. Captures what composers should do according to theoretical constraints.

Evaluation

Both models are evaluated on the same dataset of chord progressions. Metrics include harmonic coherence, adherence to voice leading rules, and subjective musical quality. A rule-based system supplements both models for error correction and constraint enforcement.

## Repository Structure

├── models/                          
│   ├── data/                        
│   │   ├── chord_bases.txt          
│   │   ├── chord_bases_no_repeats.txt        
│   │   ├── chord_bases_no_repeats_proc...    
│   │   └── chord_bases_processed.pt          
│   ├── model_states/                
│   │   ├── chord_model_epoch_10.pth
│   │   └── chord_model_no_repeats_epo...
│   └── shared_files/                
│       ├── chord_identities.py      
│       ├── generate_chords_from_model.py     
│       ├── genetic_algo.py          
│       ├── lstm.py                  
│       └── markov_model.py          
├── visualizations/                  
├── Images/                          
├── additional_notes/                
│   └── README.md
├── main.py                         
├── remove_repeats.py                
├── requirements.txt                 
└── README.md

---

Getting Started

Prerequisites

Python 3.10+
pip install -r requirements.txt

Running the Models

# Markov Chain
python markov/train.py --data data/progressions.csv
python markov/generate.py --length 8

# Genetic Algorithm
python genetic/evolve.py --generations 500 --population 100

---

Timeline

Milestone	Target Date
Project Proposal	January 25, 2025
First Check-In	Last week of February
Second Check-In	March 27th 2026
Presentation Slides	April 8, 2026
Online Presentation	April 12, 2026
2 Page Overview + Final Deliverables	April 17, 2026

Acknowledgments

We thank the CS 4100 teaching staff - in particular the TAs specializing in genetic algorithms and machine learning - for their guidance throughout this project.

---

CS 4100 - Artificial Intelligence, Northeastern University