Cognitive Architectures
Comparisons, lists, societies:

Architectures with software for experimentation:

Theoretical ideas and architectures without available software:

Projects related to cognitive architectures (mostly BICA), some already completed:
  • DARPA Completed programs: Biologically-Inspired Cognitive Architectures (BICA), Learning Applied to Ground Robots (LAGR), Real-World Reasoning (REAL) and more.
  • Architecture of Brain and Mind, integrating high level cognitive processes with brain mechanisms and functions in a working robot.|
  • Blue Brain, EPFL+IBM, 10K Hodgkin-Huxley compartmental neurons (at the ion channels details) to model neocortical column, using NEURON simulator, completed in 2008.
  • Brain-i-Nets, Novel Brain-Inspired Learning Paradigms for Large-Scale Neuronal Networks, EU project (2010-12).
  • BrainScales, Brain-inspired multiscale computation in neuromorphic hybrid systems, EU project (2011) with 18 groups, wafer scale integration, in-vivo biological experimentation and computational analysis, also will test probabilistic inference mechanisms.
  • C2S2, Cognitive Computing via Synaptronics and Supercomputing, part of DARPA Synapse initiative; initial simulations for 1.6G single compartment spiking neurons, ~9T synapses on about 150K CPUs of IBM Dawn Blue Gene/P supercomputer, newer work is on hardware chips.
  • Digital Aristotle - project Halo.
  • Semantic Inferencing on Large Knowledge (SILK)
  • D-Omar in Java, an agent development environment.
  • Emergent - extension of PDP++,Randall C. O'Reilly, BICA architectures are build around it.
  • Evolved Machines, growing large neural nets for perception!
  • FACETS, Fast Analog Computing with Emergent Transient States, EU project with 15 groups, neuromorphic analog spiking 256 neuron chips with 128K synapses, STDP learning, finished in 2010, followed by BrainScales.
  • Keenan's Software for Memory and Cognition, psychological experiments
  • NeuroEvolution of Augmenting Topologies (NEAT) and HyperNEAT
  • Nengo rapid spiking network simulator
  • Neurogrid, Brains in Silicon at Stanford University, CMOS chips with 64K neurons x 16 on a board, Integrated and Fire neurons with additional simplified somatic compartments, about 300 transistors/neuron used, analog communication, 6G connections, real time, 5 W power only.
  • OpenNERO research and education platform with graphical display of a 3-D physical world that includes multiple agents with embedded sensors and effectors and multiple objects; tutorial on neuroevolution and application in real-time games.
  • Memory-based prediction and Neocortex project |
  • SpiNNaker, A Universal Spiking Neural Network Architecture, Manchester; hardware chips, 1000 spiking point neurons each, based on ARM processors; 64K chips + 18 CPU cores architecture finished in 2012.

See also the most ambitious AI projects.
Comparison of Artificial Brain Projects.



Last changes: 2.07.2013 Wlodzislaw Duch