A computational framework for connection matrix theory (Q2239809)

An important task in topological data analysis is the construction of a filtered simplicial complex \(X\) and its associated chain complex \(C(X)\) from the data, and the computation of its homology. Typically these complexes are huge and costly to handle. In the present paper the authors introduce an algorithm to transform a graded chain complex so that the transformed complex, called Conley complex, has a simple boundary operator, called connection matrix. This is done in a categorical context: the original complex and the transformed one are equivalent in a certain homotopy category of graded chain complexes. The authors relate their approach to the original definitions of connection matrices in the context of Conley index theory due to \textit{R. D. Franzosa} [Trans. Am. Math. Soc. 311, No. 2, 561--592 (1989; Zbl 0689.58030)] and \textit{J. W. Robbin} and \textit{D. A. Salamon} [Ergodic Theory Dyn. Syst. 12, No. 1, 153--183 (1992; Zbl 0737.58033)]. The paper also contains a pseudo-code for the computation of connection matrices, and a reference to the implementation of the algorithm by the first and third authors. Moreover the authors show that the Conley complex and its connection matrix can be used to compute the persistent homology of a filtered chain complex.