Regressor

The Regressor process provides real-time machine learning regression capabilities using RapidLib algorithms. Unlike classification which predicts discrete classes, regression predicts continuous values - perfect for mapping complex sensor data to smooth control parameters.

It is very close to the Classifier process: the difference is that Regressor interpolates while Classifier outputs the exact input value that was set during training.

Overview

The Regressor enables:

• Multi-dimensional input to output mapping
• Smooth interpolation between training points
• Real-time continuous value prediction
• Non-linear relationship learning
• Adaptive parameter control

Perfect for:

• Gesture-to-sound mapping
• Environmental sensor control
• Complex motion tracking
• Adaptive audio/visual parameters
• Smooth interactive responses

Inputs

Outputs

Usage Workflow

1. Define Mapping

Decide what you want to map:

• Inputs : Sensor values, analysis data, control positions
• Outputs : Audio parameters, visual controls, device values

2. Data Collection

For each training point:

1. Set inputs to desired state
2. Set Target to desired outputs
3. Trigger Record
4. Repeat across your mapping space

3. Training and Use

Training Phase:
[Input State] → [Set Target] → [Record] → [Collect Examples] → [Train]

Runtime Phase:
[Live Input] → [Regressor] → [Smooth Output] → [Parameter Control]

Example Applications

Gesture-to-Sound Mapping

Map hand position to audio parameters:

[Hand Tracker] → [X, Y Position] → [Regressor] → [Volume, Pitch, Filter]
                                                        ↓
                                                [Audio Synthesis]

Training process:

1. Move hand to top-left: Record [high volume, low pitch, open filter]
2. Move to bottom-right: Record [low volume, high pitch, closed filter]
3. Add more points to define the mapping space
4. Train the model for smooth interpolation

Environmental Control

Map multiple sensors to lighting parameters:

[Temperature] → [Feature Vector] → [Regressor] → [Hue, Saturation, Brightness]
[Humidity]    →                                             ↓
[Light Level] →                                      [LED Controller]
[Occupancy]   →

Expressive Instrument

Create nuanced musical control:

[Pressure Sensors] → [Regressor] → [Synthesis Parameters]
[Tilt Sensors]     →                      ↓
[Proximity]        →              [Multi-dimensional Control]

Advanced Mapping Techniques

Multi-dimensional Spaces

Create complex mappings with multiple inputs and outputs:

Input Space: [x, y, pressure, tilt] Output Space: [volume, pitch, timbre, reverb, delay]

This creates a 4D → 5D mapping for rich expressive control.

Temporal Features

Include time-based features:

[Current Values] → [Feature Vector] → [Regressor] → [Predicted Next Values]
[Rate of Change] →
[Acceleration]   →
[History Window] →

Hierarchical Mapping

Use multiple regressors for different aspects:

[Gesture Data] → [Regressor 1] → [Musical Structure]
               → [Regressor 2] → [Timbral Control]  
               → [Regressor 3] → [Spatial Position]

Training Strategies

Sparse Training

Efficiently cover the input space:

1. Corner Points : Train at extremes first
2. Grid Sampling : Regular samples across space
3. Important Regions : More points where behavior changes rapidly
4. Edge Cases : Handle boundary conditions

Iterative Refinement

Improve the mapping over time:

1. Initial Training : Basic mapping with key points
2. Test and Evaluate : Find areas needing improvement
3. Targeted Collection : Add examples where needed
4. Retrain : Update model with new data

Quality Control

Ensure good training data:

• Consistency : Similar inputs should have similar outputs
• Coverage : Sample across the full input range
• Smoothness : Avoid sudden jumps in output values
• Validation : Test with holdout data

Performance Optimization

Real-time Considerations

For live performance:

• Latency : Neural networks add minimal latency
• Update Rate : Match prediction rate to control needs
• Stability : Ensure smooth outputs during rapid input changes

Data Management

Efficient training:

• Dataset Size : More data generally improves quality
• Memory Usage : Large datasets consume more memory
• Training Time : Neural networks take longer to train
• Model Storage : Save trained models for reuse

Integration Examples

Interactive Art Installation

[Visitor Position] → [Regressor] → [Ambience Parameters]
[Movement Speed]   →                    ↓
[Group Size]       →              [Audio Layers]
[Time of Day]      →              [Visual Intensity]
                                  [Color Temperature]

Adaptive Music System

[Audio Analysis] → [Regressor] → [Generative Parameters]
[User Behavior]  →                     ↓
[Context Data]   →              [Melody Complexity]
                                [Harmonic Density]
                                [Rhythmic Variation]

Responsive Architecture

[Environmental Sensors] → [Regressor] → [Building Response]
[Occupancy Patterns]    →                     ↓
[Energy Usage]          →              [Lighting Levels]
[External Conditions]   →              [Ventilation Rate]
                                      [Space Configuration]

Best Practices

Data Collection

1. Plan Your Space : Map out input/output relationships beforehand
2. Systematic Sampling : Use structured approaches to cover space
3. Test As You Go : Validate mappings during collection
4. Document Examples : Keep notes on training points

Related Processes

• Classifier - For discrete category prediction
• Interpolator - Simple linear interpolation
• Mapping - Basic parameter mapping
• Analysis - Feature extraction

Try it!

Try it by downloading this simple example!