This talk discusses how density based supervised and unsupervised machine learning in the form of classification and clustering can be performed with sparse grids which are built using the combination technique. Using classification an algorithm can learn which points of an input set are more likely to accept a certain class to then label unknown new data accordingly. Clustering on the other hand lets an algorithm label a completely unknown input set based on similar properties between points. To accomplish that, a 'nearest-neighbor-graph' connecting most of the set is built and then reduced according to the estimated density, constructing new clusters. Such high-dimensional machine learning tasks are usually infeasible to compute with full grid density estimation, so a sparse grid based approach is utilized to ease the so called 'curse of dimensionality' that arises with high dimensional computations.

Keywords: Machine Learning, Classification, Clustering, Density Estimation, Sparse Grids, Combination Technique, Curse of Dimensionality