A modern approach to the study of large, high-dimensional data sets, applying methods from algebraic topology. The approach is robust to noise. Its motivation is to rigorously study the ‘shape’ of the data; the main tool is called persistent homology.