paper reading

Variational Implicit Processes . (2019, May)

Ma, C., Li, Y., & Hernández-Lobato, J. M. (2019, May). Variational implicit processes. In International Conference on Machine Learning (pp. 4222-4233).

IP 定义：

In contrast to prescribed probabilistic models (Diggle & Gratton , 1984 ) that assign explicit densities to

possible outcomes of the model, implicit models implicitly assign probability measures by the specification of the data generating process .

IP没有事先定义的probabilisitic model 通过数据 implicitly assign. i..e. GAN

隱函数（implicit function）

Similar to the construction of Gaussian processes (GPs), an implicit process (IP) assigns implicit distributions over any finite collections

of random variables. Therefore IPs can be much more flexible than GPs when complicated models like neural networks are used for the implicit distributions.

和GP类似，IP的隐分布也是通过一个有限的rv集合定义。因为ips可以利用更加复杂模型如NN定义所以更加灵活

With an IP as the prior, we can directly perform (variational) posterior inference over functions in a non-parametric fashion.

Deep Gaussian processes for regression using approximate expectation propagation 2016

Bui, T., Hernández-Lobato, D., Hernandez-Lobato, J., Li, Y., & Turner, R. (2016, June). Deep Gaussian processes for regression using approximate expectation propagation. In International conference on machine learning (pp. 1472-1481)