COMPUTATION BY SYMMETRY OPERATIONS IN A STRUCTURED MODEL OF THE BRAIN - RECOGNITION OF ROTATIONAL INVARIANCE AND TIME-REVERSAL

被引:17
作者
MCGRANN, JV
SHAW, GL
SHENOY, KV
LENG, XD
MATHEWS, RB
机构
[1] UNIV CALIF IRVINE,CTR NEUROBIOL LEARNING & MEMORY,IRVINE,CA 92717
[2] UNIV CALIF IRVINE,DEPT PHYS,IRVINE,CA 92717
来源
PHYSICAL REVIEW E | 1994年 / 49卷 / 06期
关键词
D O I
10.1103/PhysRevE.49.5830
中图分类号
O35 [流体力学]; O53 [等离子体物理学];
学科分类号
070204 ; 080103 ; 080704 ;
摘要
Symmetries have long been recognized as a vital component of physical and biological systems. What we propose here is that symmetry operations are an important feature of higher brain function and result from the spatial and temporal modularity of the cortex. These symmetry operations arise naturally in the trion model of the cortex. The trion model is a highly structured mathematical realization of the Mountcastle organizational principle [Mountcastle, in The Mindful Brain (MIT, Cambridge, 1978)] in which the cortical column is the basic neural network of the cortex and is comprised of subunit minicolumns, which are idealized as trions with three levels of firing. A columnar network of a small number of trions has a large repertoire of quasistable, periodic spatial-temporal firing magic patterns (MP's), which can be excited. The MP's are related by specific symmetries: Spatial rotation, parity, ''spin'' reversal, and time reversal as well as other ''global'' symmetry operations in this abstract internal language of the brain. These MP's can be readily enhanced (as well as inherent categories of MP's) by only a small change in connection strengths via a Hebb learning rule. Learning introduces small breaking of the symmetries in the connectivities which enables a symmetry in the patterns to be recognized in the Monte Carlo evolution of the MP's. Examples of the recognition of rotational invariance and of a time-reversed pattern are presented. We propose the possibility of building a logic device from the hardware implementation of a higher level architecture of trion cortical columns.
引用
收藏
页码:5830 / 5839
页数:10
相关论文
共 54 条