论文部分内容阅读
提出和分析了一种新型的反馈型随机神经网络 ,并将其用于解决复杂的人脸识别问题 .该模型采用随机型加权联接 ,神经元为简单的非线性处理单元 .理论分析揭示该网络模型存在唯一的收敛平稳概率分布 ,当网络中神经元个数较多时 ,平稳概率分布逼近于 Boltzmann- Gibbs分布 ,网络模型与马尔可夫随机场之间存在密切关系 .在设计了一种新型模拟退火和渐进式 Boltzmann学习算法后 ,系统被成功地应用于难度较大的静态和动态人像识别 ,实验结果证实了系统的可行性和高效率
A new type of feedback stochastic neural network is proposed and analyzed, which is used to solve complex face recognition problems. The model uses a stochastic weighted connection, and the neuron is a simple non-linear processing unit. Theoretical analysis reveals that the network When the number of neurons in the network is large, the stationary probability distribution approaches the Boltzmann-Gibbs distribution, and there is a close relationship between the network model and the Markov random field. When a new type of simulation Annealing and progressive Boltzmann learning algorithm, the system was successfully applied to the more difficult static and dynamic portrait recognition, the experimental results confirmed the feasibility of the system and high efficiency