Valentina Staneva will give a talk titled “Introduction to Variational Inference and its Applications”
Variational inference is an optimization-based alternative to MCMC methods for estimating posterior distributions. We will go through the steps of formulating the optimization problem and we will also discuss how its stochastic version can handle large data sets. As an example we will show how the framework is applied to inference from latent variable models which are widely used for dimensionality reduction in various domains. Time permitting I will show how to implement those using the Keras library in Python.
More information on this seminar series available here: https://escience.washington.edu/get-involved/topics-in-data-science-seminar/
This event is open to the public.