Skip to content

10,000D Vector Symbolic Architecture • XOR Quorum ECC • Counterfactual Cognition • Gemini Research Prototype

Notifications You must be signed in to change notification settings

AdaWorldAPI/dragonfly-vsa

Repository files navigation

🜲 Dragonfly VSA v0.7.0

*"Compound eye

Recent Breakthrough: Signal Separation (v0.7.3)

The Flip

We discovered that flipping the direction of the DTO dramatically improves fidelity:

Approach Direction Fidelity
Old 10KD → compress → 1024D → expand → 10KD 62.5%
New 1024D → expand → 10KD → compute → compress ~97%

The Key Insight

1024D qualia is the SOURCE OF TRUTH, not a projection.

When concepts are defined as SPARSE combinations of atoms:

  • 10 atoms: 96.85% fidelity
  • 50 atoms: 96.88% fidelity
  • 100 atoms: 96.82% fidelity

The Architecture

1024D QUALIA (Source)          10KD RESONANCE (Workspace)
─────────────────────          ─────────────────────────
 cat (~20 atoms)    ─────EXPAND────→  cat_10k (binary)
 on  (~10 atoms)    ─────EXPAND────→  on_10k (binary)
                                           │
                                      XOR BIND
                                           │
 triple (~40 atoms) ←───COMPRESS───  triple_10k (binary)

New Files

  • src/clean_qualia_dto.py — High-fidelity 1024D ↔ 10KD transformation
  • docs/signal_separation.md — Complete documentation of the breakthrough

See Signal Separation for full details.

About

10,000D Vector Symbolic Architecture • XOR Quorum ECC • Counterfactual Cognition • Gemini Research Prototype

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published