You are here

Home

Spaun

Mon, 03/19/2012 - 13:26 — xchoo
Description: 
Spaun is a biologically realistic model of cognition that is not only able to perform multiple (at least 10) cognitive, perceptual, and motor tasks, but also utilizes the same model parameters across all tasks. Spaun is able to perform tasks that encompass strictly visual tasks (e.g. recognition of handwritten digits), memory tasks (e.g. forward and backward recall of a list), simple cognitive tasks (e.g. counting), and complex fluid intelligence tasks (e.g. solving the Raven's Progressive Matrices).
Keywords: 
spaun
vision
basal ganglia
learning
reinforcement learning
inference
neural coding
Authors: 
Chris Eliasmith
Terry Stewart
Feng-Xuan Choo
Trevor Bekolay
Travis DeWolf
Charlie Tang
Daniel Rasmussen
Peer Reviewed: 
Yes
Publication: 
Eliasmith, C., et al. (2012). A large-scale model of the functioning brain. Science. 338(6111), 1202-1205.
Publication URL: 
http://nengo.ca/publications/spaunsciencepaper
Instructions: 

See videos of Spaun's performance.
Read a brief description of Spaun's architecture.

SPAUN README
============

To run spaun, follow these instructions:
0. Extract the files from the zip file.
1. Install nengo 1403 as per the readme.txt file
2. Run nengo 1403
3. Within nengo, press CTRL+P to bring up the scripting console,
and type "run run_spaun_win.py" (if you are on a windows-based
system) or "run run_spaun_lin.py" (if you are on a linux-based
system).
4. The log files that are generated can be found in spaun/out_data
located within this directory.

Notes:
--------
- This model requires a machine with at least 24GB of RAM to run the full implementation.
Estimated run times for a quad-core 2.5GHz are 3 hours per 1 second
of simulation time.
- See the run_spaun.py file in the spaun directory for experiment
options.

Model: 
spaun.zip