Continuous attractors and learning
Several recent models of learning in the CNS have referred to so-called
continuous attractors. A continuous attractor (which is actually
a set of contiguous point attractors) has the advantage that it can potentially
store all values of an analogue input. Seung (1996) theorized that eye movements
are controlled by a quasi-linear continuous attractor system, while Samsonovich
et al (1997) were possibly the first to clearly formulate the theory that
hippocampal "place cells" form a continuous attractor. Helge Malmgren's research
in the ANNIMAB group on continuous attractors concentrates on formulating
conditions for the existence of non-linear learning continuous
attractors in two-dimensional dynamical systems and to apply these findings
to memory at the single-cell level. Results have been presented at the 5th
and 6th ICCNS in Boston.
References:
Malmgren H, Representations can re-present.
Notes on unsupervised online learning in sleep-wake systems with continuous
attractors. Presentation at the Fifth International Conference on
Cognitive and Neural Systems, Boston, May 31-June 2, 2001.
Malmgren H, Conditions for forced
learning of graded responses. Poster presentation at the Sixth International
Conference on Cognitive and Neural Systems, Boston, May 30-June 1, 2002.
Pdf format.
Samsonovich A, McNaughton BL, "Path integration and cognitive
mapping in a continuous attractor neural network model". Journal of Neuroscience
17 (1997), 5900-20.
Seung HS, "How the brain keeps the eyes still". Proceedings
of the National Academy of Sciences USA 93 (1996), 13339-44.
Back to main (Swedish) ANN page
Back to main
(English) ANN page
