Abstract
The Companion cognitive architecture supports experiments in achieving human-level intelligence. The seven key features of cognitive architecture include analogical processing, extensive conceptual knowledge, flexible reasoning, and coarse-grained distributed implementation, broad learning at multiple levels, continuous operation, and natural interaction. The model for analogical matching is the Structure-Mapping Engine (SME) that computes mappings using algorithm, operating in polynomial time. The model for similarity-based retrieval is many are called/few are chosen (MAC/FAC) that takes as input a probe and a case library. Another design goal for Companions is to emulate the parallelism that's evident in human behavior. Companions are implemented as distributed systems that allocate individual nodes of a cluster computer to semi-independent, asynchronous processes (agents). Agents communicate internally using the Knowledge Query and Manipulation Language (KQML) with callbacks to support asynchronous queries and subscriptions to events.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 36-46 |
| Number of pages | 11 |
| Journal | IEEE Intelligent Systems |
| Volume | 24 |
| Issue number | 4 |
| DOIs | |
| State | Published - 2009 |
Funding
This work has been supported by DARPA under an Information Processing Techniques Office seedling and the Transfer Learning Program, and by the US Office of Naval Research under the Intelligent Systems and Cognitive Science Programs.
ASJC Scopus subject areas
- Computer Networks and Communications
- Artificial Intelligence