220 likes | 365 Views
Evolutionary Path to Biological Kernel Machines. Magnus Jändel magnus@jaendel.se. Swedish Defence Research Agency. Summary. It is comparatively easy for organisms to implement support vector machines .
E N D
Evolutionary Path to Biological Kernel Machines Magnus Jändel magnus@jaendel.se Swedish Defence Research Agency
Summary It is comparatively easy for organisms to implement support vector machines. Biological support vector machines provide efficient and cost-effective pattern recognition with one-shot learning [1]. The support vector machine hypothesis is consistent with the architecture of the olfactory system [1]. Bursts in the thalamocortical system may be related to support vector machine pattern recognition [2]. An efficient implementation reuses machinery for learning action sequences [3]. 1) Jändel, M.: A neural support vector machine. Neural Networks 23, 607-613 (2010). 2) Jändel, M.: Thalamic bursts mediate pattern recognition. Proceedings of the 4th International IEEE EMBS Conference on Neural Engineering 562–565 (2009). 3) Jändel, M.: Pattern recognition as an internalized motor programme. To appear in proc. of ICNN 2010. Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Outline Support vector machine definition Evolutionary path to a neural SVM Conclusions and olfactory model Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Maximum margin linear classification Consider binary classification with m training examples: Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Transform to high-dimensional feature space Zero-bias SVM: Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Zero-bias n-SVM Maximize: and Subject to: where Solve by iterative gradient ascent in the a-space hyperplane where The margin of the i:th example in feature space! Classification function: Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 1 SS PR Sensor system Simple hard-wired pattern recognizer Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 2 x SS SM PR Sensor system Sensory Memory Simple hard-wired pattern recognizer Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 3 x SS SM PR Sensor system Sensory Memory Simple hard-wired pattern recognizer AM Associative memory Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 4 x SS SM PR x y´ AM - Significant patterns and the associated valence are stored in the AM. - Sufficiently similar inputs make the AM recall the valence of a stored pattern. Zero-bias n-SVM Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 5 x SS SM PR x x´, y´ - Significant patterns and the associated valence are stored in the AM. - Sufficiently similar inputs make the AM recall the valence of a stored pattern - The PR modulates the recalled valence y´ with a similarity measure comparing input x with the stored pattern x´ according to, AM Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 6 x SS SM PR x xi, yi - The OM oscillates between memory states - The PR computes a weighted average over the valences of all stored examples, OM Zero-bias n-SVM Stage 6 implements the classification function of a zero-bias SVM. Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Oscillating Associative Memory Hopfield associative memory OM Model N neurons with binary output zi m memory patterns Update rule Each oscillation selects the next state with uniform probability. The average endurance time of state i is Ti. Imprint m memory patterns x(k) The probability of finding the OM in state i is, One-shot learning! Oscillating memory - Firing cell nuclei are exhausted - Active synapses are depleted Modes with perpetual oscillation between attractors. Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 7 xj feedback SS SM PR xi, yi xj - Learning feedback Bij tunes memory weights - Real-world experiments are required Bij OM xi is the present example presented by the OM xjis the sensory input yj is the valence of xj as learnt from hard-earned experience For each OM oscillation apply the learning rules, and Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Stage 8 xj SS SM PR xj xi, yi Bij xi OM - OM patterns are set up in sensory memory while sleeping - OM weights tuned in virtual experiments - No need for external feedback - Implements a zero-bias n-SVM Zero-bias n-SVM Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Learning SVM weights xj SS SM PR xj xi, yi Bij xi For each OM oscillation apply the learning rules, OM and where gives Averaging over “trapped examples” with probability distribution where Zero-bias n-SVM Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Summary of support vector machine implementation xj x SS SM PR xj xi, yi Bij xi OM Classification process Learning new training examples Learning weights of training examples Zero-bias n-SVM Research program Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010
Olfactory model PPC M2 D5 D4 OM D3 HOBS M3 D1 CL M1 Trap OB D2 AOC APC APC – Anterior piriform cortex PPC – Posterior piriform cortex AOC – Anterior olfactory cortex OB – Olfactory bulb HOBS – Higher-order brain systems Magnus Jändel, Brain Inspired Cognitive Systems, 15 July 2010