ml_quaternion_features

ML Quaternion Features

This example demonstrates using quaternions as features for machine learning, combining the quaternion and ML modules.

What You'll Learn

• Extracting features from quaternions for ML
• Using quaternion components as input features
• Training clustering models on quaternion data
• Visualizing quaternion features

Prerequisites

• V compiler installed (download here)
• VSL library installed (installation guide)
• No additional system dependencies required

Running the Example

# Navigate to this directory
cd examples/ml_quaternion_features

# Run the example
v run main.v

Expected Output

The example generates:

• Console output: Data generation, training progress, and accuracy results
• 2D Plot: Visualization of quaternion features (x vs y components)
• Clustering results: Accuracy of K-means clustering on quaternion features

Code Walkthrough

1. Generating Quaternion Data

import vsl.quaternion
import math

n_samples := 50
angle := math.pi / 4.0
mut quaternion_data := []quaternion.Quaternion{}

// Create two groups of orientations
for i in 0..n_samples {
    if i < n_samples / 2 {
        // Group 1: x-axis rotations
        q := quaternion.from_axis_anglef3(angle, 1.0, 0.0, 0.0)
        quaternion_data << q
    } else {
        // Group 2: y-axis rotations
        q := quaternion.from_axis_anglef3(angle, 0.0, 1.0, 0.0)
        quaternion_data << q
    }
}

We generate synthetic orientation data with two distinct groups.

2. Feature Extraction

import vsl.quaternion

mut feature_matrix := [][]f64{}
quaternion_data := []quaternion.Quaternion{}  // Assume populated

// Extract quaternion components as features
for q in quaternion_data {
    feature_matrix << [q.x, q.y]  // 2D for Kmeans compatibility
}

Quaternion components (w, x, y, z) become 4-dimensional feature vectors.

3. Machine Learning

import vsl.ml

feature_matrix := [][]f64{}  // Assume populated
mut data := ml.Data.from_raw_x(feature_matrix)!
mut train_data, _ := data.split(0.8)!
nb_classes := 2
mut model := ml.Kmeans.new(mut train_data, nb_classes, 'quaternion_kmeans')
model.train(epochs: 10)

We train a K-means clustering model to identify the two groups.

Use Cases

• Robotics: Classify robot orientations
• Motion Analysis: Identify movement patterns
• Sensor Data: Process IMU/orientation sensor data
• Animation: Classify animation poses

Experiment Ideas

Try modifying the example to:

• More features: Add derived features (magnitude, angle, etc.)
• Different algorithms: Try regression or classification
• Real data: Use actual sensor data
• More groups: Add a third orientation group
• Feature engineering: Create additional features from quaternions

Related Examples

• ml_kmeans - Basic K-means clustering
• quaternion_orientation_tracking - Orientation tracking
• quaternion_rotation_3d - Quaternion rotations

Related Tutorials

• Machine Learning Data Preparation
• Clustering
• Quaternion Introduction

Troubleshooting

Low accuracy: Try different initial centroids or more training epochs

Plot doesn't open: Ensure web browser is installed

Module errors: Verify VSL installation

---

Explore more ML and quaternion examples in the examples directory.