Experiments with Neural Processes, Mike Clark at PMLG

Mike Clark from Three Springs Technology gave a talk on Neural Processes in March 2020.   

These novel networks were first introduced in 2018 and are similar to Gaussian Processes in behaviour, have a meta-learning component, and have a latent space similar to variational autoencoders.

From a practical standpoint, you may be interested in this talk if you work with time series and want uncertainty estimates when you do predictions. Especially if your problem is partly periodic in nature, like weather, tides, or power consumption.

The code and experiments that this talk are based on is here: https://github.com/3springs/attentive-neural-processes

In case you missed the talk the slides are available here: https://docs.google.com/presentation/d/1xW726toi4JfS0287fv_mxXOaUX2Vlvrtw1hJ6XcjSOU/edit?usp=sharing