Mean field approach to Bayes learning in feed-forward neural networks.
Mean field approach to Bayes learning in feed-forward neural networks.
复制标题
前馈神经网络中贝叶斯学习的平均场方法。
DOI：
10.1103/physrevlett.76.1964
复制
发表时间：
1996
影响因子：
8.6
通讯作者：
Winther
中科院分区：
文献类型：
--
作者：
Opper;Winther
基于无序系统统计力学的Thouless-Anderson-Palmer平均场方法，我们提出了一种实现前馈网络贝叶斯最优预测的算法。我们推测，我们的方法在热力学极限下是精确的。该算法的结果是对预测进行简单的内置留一交叉验证。对简单感知器和委员会机器的模拟结果与复制理论的结果很好地吻合。
We propose an algorithm to realize Bayes optimal predictions for feed-forward networks which is based on the Thouless-Anderson-Palmer mean field method developed for the statistical mechanics of disordered systems. We conjecture that our approach will be exact in the thermodynamic limit. The algorithm results in a simple built-in leave-one-out cross validation of the predictions. Simulations for the case of the simple perceptron and the committee machine are in excellent agreement with the results of replica theory.