Knowledge Representations

help edit space_dashboard

Atomspace technologies form the symbolic foundation of the Hyperon neural-symbolic approach. Systems such as DAS and MORK enable a dynamic knowledge metagraph where code and data are interchangeable.

DAS (Distributed AtomSpace)

help edit space_dashboard
Responsible: André Senna

DAS is a high-speed, dynamic memory fabric for massive, mutable hypergraphs. It decouples vast persistence (Long-Term Importance) from immediate attention dynamics (Short-Term Importance in RAM), governed by an Attention Broker that constrains search space to relevant atoms.

</> Example Implementation
Distributed query
Shows how DAS queries leverage the Attention Broker to efficiently search across distributed Atomspaces.
; DAS distributed hypergraph query ; Attention Broker constrains search !(das-query (pattern (Inheritance $x Animal)) (importance-threshold 0.7) (max-results 10)) ; Returns atoms with STI > 0.7 ; matching the inheritance pattern

Technical Deep Dive: DAS Full — Attention Broker as ECAN engineering surrogate, four-layer storage taxonomy, AtomSpace-Scheme stack, and implementation findings.

MORK (MeTTa Optimized Reduction Kernel)

help edit space_dashboard
Responsible: Adam Vandervorst, Luke Peterson, Remy Clarke

MORK is an ultra-high-performance hypergraph engine achieving 1000x-1000000x speedups via optimized Trie-Map structures. Its zipper-based multi-threaded VM navigates complex reasoning paths with near-instant access, and uses WASM for seamless external code integration.

</> Example Implementation
Trie-Map traversal
Demonstrates MORK's optimized Trie-Map structure for ultra-fast hypergraph insertion and querying.
; MORK high-performance hypergraph ops ; Zipper-based navigation !(mork-insert (Edge "is-a" "Cat" "Animal")) !(mork-insert (Edge "is-a" "Dog" "Animal")) !(mork-query (Edge "is-a" $x "Animal")) ; => ["Cat", "Dog"] ; 1000x faster than naive graph search

Technical Deep Dive: MORK Full — PathMap formalism, selectivity theorem, ZAM execution model, four-layer architecture, known limitations, and implementation findings.