COMPETITION AND MULTIPLE CAUSE MODELS

被引:44
作者
DAYAN, P [1 ]
ZEMEL, RS [1 ]
机构
[1] SALK INST, COMPUTAT NEUROBIOL LAB, SAN DIEGO, CA 92186 USA
关键词
D O I
10.1162/neco.1995.7.3.565
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
If different causes can interact on any occasion to generate a set of patterns, then systems modeling the generation have to model the interaction too. We discuss a way of combining multiple causes that is based on the Integrated Segmentation and Recognition architecture of Keeler et al. (1991). It is more cooperative than the-scheme embodied in the mixture of experts architecture, which insists that just one cause generate each output, and more competitive than the noisy-or combination function, which was recently suggested by Saund (1994a,b). Simulations confirm its efficacy.
引用
收藏
页码:565 / 579
页数:15
相关论文
共 16 条
  • [1] Finding Minimum Entropy Codes
    Barlow, H. B.
    Kaushal, T. P.
    Mitchison, G. J.
    [J]. NEURAL COMPUTATION, 1989, 1 (03) : 412 - 423
  • [2] Barlow H. B., 1961, P331
  • [3] DAYAN P, 1995, IN PRESS NEURAL COMP
  • [4] FORMING SPARSE REPRESENTATIONS BY LOCAL ANTI-HEBBIAN LEARNING
    FOLDIAK, P
    [J]. BIOLOGICAL CYBERNETICS, 1990, 64 (02) : 165 - 170
  • [5] Hinton GE., 1994, ADV NEURAL INFORM PR, V6, P3, DOI DOI 10.1021/jp906511z
  • [6] TASK DECOMPOSITION THROUGH COMPETITION IN A MODULAR CONNECTIONIST ARCHITECTURE - THE WHAT AND WHERE VISION TASKS
    JACOBS, RA
    JORDAN, MI
    BARTO, AG
    [J]. COGNITIVE SCIENCE, 1991, 15 (02) : 219 - 250
  • [7] Adaptive Mixtures of Local Experts
    Jacobs, Robert A.
    Jordan, Michael I.
    Nowlan, Steven J.
    Hinton, Geoffrey E.
    [J]. NEURAL COMPUTATION, 1991, 3 (01) : 79 - 87
  • [8] KEELER J, 1991, ADV NEURAL INFORMATI, V3, P557
  • [9] NOWLAN SJ, 1990, CRGTR905 U TOR DEP C
  • [10] NOWLAN SJ, 1993, ADV NEURAL INFORMATI, V5, P369