Concepts¶
Understanding HyperBinder's core concepts will help you use it effectively.
Key Ideas¶
Neurosymbolic Computing¶
HyperBinder combines two AI paradigms:
- Symbolic AI: Logic, rules, exact matching (like SQL or graph databases)
- Neural AI: Embeddings, similarity, semantic understanding
Most queries blend both - find similar items, then filter by exact criteria.
Hyperdimensional Computing (HDC)¶
Unlike traditional vector databases that treat embeddings as opaque blobs, HyperBinder uses Hyperdimensional Computing - a compositional encoding system where:
- Structured data is encoded preserving relationships
- Components can be extracted from composed structures
- Sets can be represented in single vectors
- Similarity can be measured between structures
This enables queries that decompose structure, not just measure similarity.
The Compose System¶
HyperBinder's schema system uses a chemistry metaphor:
| Layer | What it is | Examples |
|---|---|---|
| Molecules | Composable structures | Pair, Triple, Bundle |
| Compounds | Domain-specific templates | KnowledgeGraph, TimeSeries |
Users typically define schemas using Molecules or Compounds.
Topics¶
| Topic | Description |
|---|---|
| The Compose System | Deep dive into Molecules and Compounds |
| Molecules vs Compounds | When to use each type of schema |
| Embeddings | Choosing and configuring embedding models |
| Intersections | Connecting collections with cross-collection queries |