生成性深度结构，用于描述数据的高阶相关特性，或其观测数据和相应类别的联

Generative depth structure, used to describe the higher-order correlation characteristics of data, or the joint probability distribution of its observation data and corresponding categories. Many deep networks can generate models by sampling effectively from the network, such as RBM, DBN, DBM. Among them, the energy-based model is the most common, the deep belief network (DBN), which can estimate both the prior probability and the posterior probability.

生成性深度结构，用于描述数据的高阶相关特性，或其观测数据和相应类别的联合 概率分布。许多深层网络可以通过从网络中采样有效的生成样本，例如 RBM, DBN, DBM, 从而生成模型。其中，基于能量的模型最常见, 深度信念网络(DBN),既可以对先验概率进行估计也可以估计后验概率。

Generative depth structure is used to describe the high-order correlation characteristics of data, or the joint probability distribution of observation data and corresponding categories. Many deep networks can generate models by effectively sampling samples from the network, such as RBM, DBN, DBM. Among them, the most common model based on energy is the deep belief network (DBN), which can estimate both the prior probability and the posterior probability.<br>