| A recurrent stochastic binary network |
| |
| 引用本文： | 赵杰煜. A recurrent stochastic binary network[J]. 中国科学F辑(英文版), 2001, 44(5): 376-388. DOI: 10.1007/BF02714740 |
| |
| 作者姓名： | 赵杰煜 |
| |
| 基金项目： | This work was supported by the National Natural Science Foundation of China (Grant No. NSFC69805002)，the Natural Science Foundation of Zhejiang Province, the Ministry of Eductation Grant for Excellent Youth，the Ningbo Science Foundation for Youth (G |
| |
| 摘 要： | Stochastic neural networks are usually built by introducing random fluctuations into the network. A natural method is to use stochastic connections rather than stochastic activation functions. We propose a new model in which each neuron has very simple functionality but all the connections are stochastic. It is shown that the stationary distribution of the network uniquely exists and it is approxi-mately a Boltzmann-Gibbs distribution. The relationship between the model and the Markov random field is discussed. New techniques to implement simulated annealing and Boltzmann learning are pro-posed. Simulation results on the graph bisection problem and image recognition show that the network is powerful enough to solve real world problems.
|
| 收稿时间： | 2001-01-18 |
A recurrent stochastic binary network |
| |
| Zhao Jieyu. A recurrent stochastic binary network[J]. Science in China(Information Sciences), 2001, 44(5): 376-388. DOI: 10.1007/BF02714740 |
| |
| Authors: | Zhao Jieyu |
| |
| Affiliation: | The IST Research Institute,Ningbo University,Ningbo 315211,China |
| |
| Abstract: | Stochastic neural networks are usually built by introducing random fluctuations into the network. A natural method is to use stochastic connections rather than stochastic activation functions. We propose a new model in which each neuron has very simple functionality but all the connections are stochastic. It is shown that the stationary distribution of the network uniquely exists and it is approxi-mately a Boltzmann-Gibbs distribution. The relationship between the model and the Markov random field is discussed. New techniques to implement simulated annealing and Boltzmann learning are pro-posed. Simulation results on the graph bisection problem and image recognition show that the network is powerful enough to solve real world problems. |
| |
| Keywords: | recurrent stochastic binary network incremental Boltzmann learning Markov random field stimulated annealing. |
| 本文献已被 CNKI 万方数据 SpringerLink 等数据库收录！ |
