Guitar Level Classifier - Neural Network

A TensorFlow.js neural network that classifies guitar players into skill levels (beginner, intermediate, advanced) based on their technical knowledge.

Part of the postgraduate project in AI Applied Software Engineering.

---

Overview

This application demonstrates a practical implementation of machine learning in Node.js, using TensorFlow.js to build, train, and execute a classification model. The neural network analyzes guitar players' technical characteristics and predicts their skill level.

---

Technical Stack

Technology	Purpose
TensorFlow.js (v4.22)	Neural network framework
Node.js	Runtime environment
Docker + DevContainer	Reproducible development environment

---

Model ArchitectureInput Layer (9 features)

↓ Dense Layer (100 units, ReLU activation) ↓ Output Layer (3 units, Softmax activation) Features (One-Hot Encoded):

• Chords: basic chords, CAGED, inverted chords
• Scales: pentatonic, natural minor, harmonic minor
• Licks: single shape, various shapes, various tones

Classification Output:

• Beginner
• Intermediate
• Advanced

---

Training Configuration

Parameter	Value
Optimizer	Adam
Loss Function	Categorical Crossentropy
Epochs	100
Shuffle	Enabled
Metrics	Accuracy

---

Quick Start

Prerequisites

• Node.js 18+
• Docker (optional, for containerized environment)

Running Locally

# Install dependencies
npm install

# Run the application
npm startRunning with DevContainerOpen the project in VS Code with Dev Containers extension and select "Reopen in Container".

Example output:
beginner (99.87%)
intermediate (0.08%)
advanced (0.05%)

Project Structure
├── index.js              # Main application (train + predict)
├── package.json          # Dependencies and scripts
├── Dockerfile            # Container image definition
├── docker-compose.yml    # Docker Compose configuration
└── .devcontainer/        # VS Code DevContainer setup
    └── devcontainer.json

Neural network implementation with TensorFlow.js in Node.js
One-hot encoding for categorical features
Softmax activation for multi-class classification
Model training with real-time epoch logging
Containerized ML development environment

To add new training samples:
const tensorLevels = [
  [1, 0, 0, 1, 0, 0, 1, 0, 0], // beginner
  [1, 1, 0, 1, 1, 0, 1, 1, 0], // intermediate
  [1, 1, 1, 1, 1, 1, 1, 1, 1], // advanced
  // Add more samples here
];

const tensorLabels = [
  [1, 0, 0], // beginner
  [0, 1, 0], // intermediate
  [0, 0, 1], // advanced
  // Add corresponding labels
];