TY - GEN
T1 - A theory of the brain - The brain uses both distributed and localist (symbolic) representation
AU - Roy, Asim
PY - 2011/10/24
Y1 - 2011/10/24
N2 - The issue of whether objects and concepts are represented in the brain by single neurons or multiple ones, where the multiple ones are conceived to represent subconcepts or microfeatures, has plagued brain-related sciences for decades, spawning different scientific fields such as artificial intelligence (AI) and connectionism. It is also a source of dispute within some of these scientific fields. In connectionism, for example, there is never ending debate between the theories of localist (in a sense symbolic) and distributed representation. To resolve this conflict, we analyze a highly publicized class of models used by connectionists (distributed representation theorists) for complex cognitive processes and show that, contrary to their claim, they actually depend on localist (symbolic) representation of higher-level concepts in these models. We also find that these connectionist models use processes similar to symbolic computation. Based on this analysis and the accumulating evidence from single-unit recordings in neurophysiology that shows that single cells can indeed encode information about single objects (e.g. a Jennifer Aniston cell in our brains), we propose the theory that the brain uses both forms of representation, localist and distributed, and that both forms may be necessary, depending on the context. Our other conjecture is that the brain uses both forms of computation, symbolic and distributed (parallel). This theory should finally resolve the decades long conflict about representation and computational processes that has generated divisions within our fields and has stalled our progress towards creating brain-like learning systems.
AB - The issue of whether objects and concepts are represented in the brain by single neurons or multiple ones, where the multiple ones are conceived to represent subconcepts or microfeatures, has plagued brain-related sciences for decades, spawning different scientific fields such as artificial intelligence (AI) and connectionism. It is also a source of dispute within some of these scientific fields. In connectionism, for example, there is never ending debate between the theories of localist (in a sense symbolic) and distributed representation. To resolve this conflict, we analyze a highly publicized class of models used by connectionists (distributed representation theorists) for complex cognitive processes and show that, contrary to their claim, they actually depend on localist (symbolic) representation of higher-level concepts in these models. We also find that these connectionist models use processes similar to symbolic computation. Based on this analysis and the accumulating evidence from single-unit recordings in neurophysiology that shows that single cells can indeed encode information about single objects (e.g. a Jennifer Aniston cell in our brains), we propose the theory that the brain uses both forms of representation, localist and distributed, and that both forms may be necessary, depending on the context. Our other conjecture is that the brain uses both forms of computation, symbolic and distributed (parallel). This theory should finally resolve the decades long conflict about representation and computational processes that has generated divisions within our fields and has stalled our progress towards creating brain-like learning systems.
UR - http://www.scopus.com/inward/record.url?scp=80054745721&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80054745721&partnerID=8YFLogxK
U2 - 10.1109/IJCNN.2011.6033224
DO - 10.1109/IJCNN.2011.6033224
M3 - Conference contribution
AN - SCOPUS:80054745721
SN - 9781457710865
T3 - Proceedings of the International Joint Conference on Neural Networks
SP - 215
EP - 221
BT - 2011 International Joint Conference on Neural Networks, IJCNN 2011 - Final Program
T2 - 2011 International Joint Conference on Neural Network, IJCNN 2011
Y2 - 31 July 2011 through 5 August 2011
ER -