PhD student Ben Shires will be presenting his work on SHEAP - Stochastic Hyperspace Embedding And Projection - at the Machine Learning Discussion Group in the Cavendish Laboratory.
Abstract. We present SHEAP - Stochastic Hyperspace Embedding And Projection – a tool designed to process the structural data obtained from a materials structure search, in order to produce a visualisation of the energy surface being sampled. We have drawn inspiration from state-of-the-art algorithms for dimensionality reduction of high-dimensional data, such as t-SNE and UMAP . We illustrate the power of SHEAP through its application to the model energy landscapes defined by systems of particles interacting via a simple Lennard-Jones pair potential.
Talks.cam link here