Swift Student Challenge 2026 Winner

Making Independent Component Analysis intuitive, visual, and playable.

Convergence transforms ICA from abstract matrix formalism into an audiovisual interaction journey. Rotate a procedural 3D sphere, manipulate eigenvector nodes, and hear hidden signals emerge in real time.

Download Public Beta

Design Intent

To connect science, software, and human perception.

The project is designed for two audiences: curious users intimidated by scientific complexity, and beginner signal-processing students who need a practical mental model before theory.

It uses real FastICA workflow and preserves mathematical structure, while translating each stage into interactions that feel tactile, legible, and emotionally engaging.

FastICA Procedural 3D Sphere Audio-Visual Sync Playable ICA
Convergence in-game ICA scene

Gameplay Video

Watch Full Game Playthrough

Watch the full Convergence gameplay flow in one continuous run, from early concepts to final signal extraction.

Visual Notes

Mathematical intuition layers used inside the experience.

Signal mixture diagram sketch
Signal Mixture Topology
Central limit theorem intuition sketch
CLT Intuition
FastICA iterative update equations
FastICA Iteration Rules

Game preview

A six-stage interactive arc

Convergence consists of six stages, from the initial idea of extracting signal from the mixed soundsource, represented by the "gaussian planet", to the final resolution stage where signals are fully extracted and converge into a song that's written for your curiousity

Phase 01

Central Limit Theorem

Phase 1 introduces the Central Limit Theorem and normal distribution through the gaussian planet, establishing why mixed soundsources naturally converge toward structured statistical behavior.

Convergence phase 1 Central Limit Theorem scene
Convergence phase 2 ICA concept introduction

Phase 02

Introducing ICA

Phase 2 introduces the Independent Component Analysis concept and visual intuition, connecting signal geometry to a playable interaction language.

Phase 03

Whitening

Phase 3 performs whitening, reshaping covariance so each axis becomes standardized and ready for iterative extraction.

Convergence phase 3 whitening process
Convergence phase 4 FastICA optimization

Phase 04

FastICA

Phase 4 runs FastICA iterative updates, where eigenvector trajectories stabilize and independent components begin to emerge from the latent space.

Phase 05

Source Extraction

Phase 5 reaches source extraction, where mixed sounds are separated into clear layers and users can hear the model's success in real time.

Convergence phase 5 source extraction result
Convergence phase 6 onto your next journey

Phase 06

Onto Your Next Journey

Phase 6 concludes the full journey with resolved signals and a musical convergence, inviting users to continue exploring signal processing beyond the game.

Accessibility

Designed For Everyone

Accessibility is integrated directly into gameplay settings, including VoiceOver labels and hints, color-aware contrast design, and Reduce Motion support for smoother comfort across users.

BUILT USING SWIFT PLAYGROUND

Tech Specs

Convergence is optimized for Apple products by using Swift/SwiftUI and utlized core functions such as AVFoundation to provide smooth experience with less system resource consumptions. Any iPad that satisfies minimum compatibiltiy would be able to run the game smoothly.

  • Target Platform: iPadOS
  • Compatibility: iPadOS 16 or later
  • Size: 16.5MB