Skip to content

Improve interactive diffeomorphism visualization - #5

Merged
microprediction merged 1 commit into
mainfrom
feature/improve-diffeomorphism-viz
May 24, 2026
Merged

Improve interactive diffeomorphism visualization#5
microprediction merged 1 commit into
mainfrom
feature/improve-diffeomorphism-viz

Conversation

@microprediction

Copy link
Copy Markdown
Owner

Summary

Major improvements to the interactive cube-to-simplex mapping visualization that resolve clustering issues and create beautiful, mathematically correct demonstrations of the Thurstone racing diffeomorphism.

Key Improvements

🔧 Core Mapping Fixes

  • Inverse Normal CDF Mapping: Replace sigmoid with inverse normal CDF for more natural ability distribution
  • Fixed Barycentric Coordinates: Corrected coordinate conversion that was mapping triangle vertices incorrectly
  • Edge-Focused Sampling: Sample cube in inverse normal space to concentrate points near boundaries
  • Deterministic Pseudo-Noise: Added smooth variation to special horse for consistent visualization

🎛️ Interface Simplification

  • Reduced Parameters: From 6 sigmoid controls to 2 symmetric parameters (scale + shift)
  • Smart Defaults: Set mean-adaptive mode and optimal parameter values as defaults
  • Cleaner UI: Symmetric parameter handling for easier experimentation

📊 Technical Improvements

  • Coordinate System Fix: Resolved critical bug causing point clustering in corner
  • Proper Racing Dynamics: Temperature-scaled softmax implementation
  • Balanced Competition: Mean-adaptive mode creates dynamic special horse competition

Visual Results

The visualization now properly demonstrates:

  • ✅ Natural point distribution across entire triangle
  • ✅ Edge concentration in cube mapping to simplex coverage
  • ✅ Beautiful mathematical demonstration of diffeomorphism properties

Testing

  • Interactive visualization loads with optimal defaults
  • Parameter adjustments produce expected changes
  • Points properly distributed across simplex
  • Documentation updated

Documentation

  • Updated main README to highlight interactive visualizations
  • Improved parameter descriptions and defaults

🚀 The diffeomorphism explorer now provides an engaging, mathematically accurate demonstration of cube-to-simplex mappings using Thurstone racing models.

🤖 Generated with Claude Code

Major fixes and enhancements to the cube-to-simplex mapping visualization:

## Core Mapping Improvements
- Replace sigmoid with inverse normal CDF mapping for better ability distribution
- Fix barycentric coordinate conversion (was mapping vertices incorrectly)
- Implement inverse normal space sampling to concentrate points near cube edges
- Add deterministic pseudo-noise to special horse for smooth variation

## Interface Simplification
- Reduce parameters from 6 sigmoid controls to 2 symmetric parameters (scale, shift)
- Set mean-adaptive as default mode (creates better competitive balance)
- Update default parameters to recommended values for optimal visualization

## Technical Fixes
- Fix coordinate system bug that caused all points to cluster in corner
- Proper temperature-scaled softmax for racing dynamics
- Symmetric parameter handling for cleaner interface

## Documentation
- Update README to mention interactive visualizations
- Improved parameter defaults create beautiful visualization on page load

The visualization now properly demonstrates cube-to-simplex diffeomorphisms
with natural point distribution across the entire triangle.

Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
@microprediction
microprediction merged commit 27fbf47 into main May 24, 2026
3 checks passed
@microprediction
microprediction deleted the feature/improve-diffeomorphism-viz branch May 24, 2026 21:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant