Doubly Stochastic Variational Inference for Deep Gaussian Processes

New Environment Test

Succeeded with Mac M1 with TensorFlow=2.13, GPflow=2.2;

Trials
- whiten=True;
- Identity mean function at the final layer;
- identity prior mean function for inducing inputs $\mathbf{Z}^{(l-1)}$ at the $l$-th intermediate layer, whose form is the same as the prior mean of outputs $\mathbf{F}^{(l)}: \mathbb{E}[\mathbf{F}^{(l)}] = m(\mathbf{F}^{(l-1)})=\mathbf{F}^{(l-1)}$.
from the forked:

🤿 Implementation of doubly stochastic deep Gaussian Process using GPflow 2.0 and TensorFlow 2.0.

Heavily based on a previous implementation of Doubly-Stochastic-DGP and the paper

@inproceedings{salimbeni2017doubly, 
  title={Doubly stochastic variational inference for deep gaussian processes}, 
  author={Salimbeni, Hugh and Deisenroth, Marc}, 
  booktitle={Advances in Neural Information Processing Systems}, 
  year={2017} 
}

Includes demos for the step function and MNIST data set.

shixinxing/DSVI-Deep-GPs

Doubly Stochastic Variational Inference for Deep Gaussian Processes