注意:tensorflow api 在 1.1.0 以后迎来重大变化,edward 的稳定版依赖于 tensorflow 1.1.0。
edward
是一个支持概率建模、推断的 Python 第三方库,官网地址:A library for probabilistic modeling, inference, and criticism.,其教程 edward tutorials。
其主要实现和支持如下三方面:
- modeling:
- directed graphical models,有向图模型;
- neural networks(基于 keras、tensorflow slim)
- implicit generative models:
- Bayesian nonparametrics & probabilistic program
- inference:
- variational inference
- Black box variational inference
- Stochastic variational inference
- Generative adversarial networks
- Maximum a posteriori estimation
- 蒙特卡洛:
- 吉布斯采样;
- 汉密尔顿蒙特卡洛
- Compositions of inference
- Expectation-Maximization
- Pseudo-marginal and ABC methods
- Message passing algorithms
- variational inference
- criticism:
- Point-based evaluations
- Posterior predictive checks