MGSC Website: http://math.uci.edu/~mgsc/
Data visualization often involves algorithms generating 2-dimensional representations of data to illustrate relationships between different data points. The standard algorithm used for data visualization is tSNE, a non-convex method that represents the differences between data points as weighted probabilities in high and low dimensions and then minimizes the "distance" between these distributions using the Kullback-Leibler divergence. Despite its wide success, there is still very little mathematical understanding of the algorithm. In this talk, we discuss tSNE and our attempt to better mathematically represent the algorithm in both theory and practice.
Kat is a 4th year graduate student working on machine learning and probability and machine learning
Kat's advisor is Roman Vershynin.
None
Pizza will be served after the talk.