Slice sampling - Discussion
成果类型：
Editorial Material
署名作者：
Downs, OB
署名单位：
Microsoft
刊物名称：
ANNALS OF STATISTICS
ISSN/ISSBN：
0090-5364
发表日期：
2003
页码：
743-748
关键词：
摘要：
The nonnegative Boltzmann machine (NNBM) is a recurrent neural network model that can describe multimodal nonnegative data. Application of maximum likelihood estimation to this model gives a learning rule that is analogous to that of the binary Boltzmann machine. While the model itself is analytically intractable an efficient stochastic version of the learning rule can be obtained using reflective slice sampling, since the slice boundaries can be determined analytically from the model. We compare this with the use of advanced mean field theory to learn a generative model for face image data.