Support Vector Machine Models of Brain Processes

Summary of the Research Program

Support vector machines are state of the art pattern recognition algorithms based on sound mathematical principles but with little apparent relation to the brain. I am working on a hunch that support vector machines are implemented in the brain. Pattern recognition in low-level perception is, in particular, a promising area of investigation.

Progress so far

The main result is a model for how a specific support vector machine (Zero-bias ν-SVM) can be implemented in neural networks [1]. This model is applied to low-level perception in the olfactory (odour recognition) system. A recent result is that neural machinery for learning action sequences have dual use for neural SVM [2].

I have also investigated how the model maps to the dynamics of the main information router in the brain - the thalamus. The support vector machine approach appears to explain the function of burst signaling in the thalamus [3]. Any complex machinery in the brain must have evolved from simpler structures. A recent manuscript [4] describes an evolutionary path for neural support vector machines.

A related research track is the investigation of multiagent systems where indivual agents employ support vector machine methods. Two papers [5] [6] describe the properties and applications of cooperating support vector machine agents. It is intriguing to consider the application of this to neuroscience.

This research is supported by The Swedish Foundation for Strategic Research.

References

  1. M. Jändel, A Neural Support Vector Machine, Neural Networks, 23 (2010) 607-613 .
  2. M. Jändel, Pattern Recognition as an Internalized Motor Programme, to appear in Proc. International Conference on Neural Networks, Amsterdam, September 2010.
  3. M. Jändel, Thalamic Bursts Mediate Pattern Recognition, Proceedings of the 4th International IEEE EMBS Conference on Neural Engineering, 562-565, 2009.
  4. M. Jändel, Evolutionary Path to Biological Kernel Machines, to appear in Proc. Brain Inspired Cognitive Systems, Madrid, July, 2010.
  5. M. Jändel, Cooperating Classifiers, International Workshop on Nature Inspired Cooperatives Strategies for Optimization 2008, published in Studies in Computational Intelligence, Vol. 236, 213-225, Springer Berlin, 2009.
  6. M. Jändel and M. Elahi, Tribal Taste: Mobile Multiagent Recommender System, Proceedings of the International Conference on Intelligent User Interfaces, 489-490, ACM New York, 2009.


This page was last modified 5 July 2010