We present a graph-based approach to support case vs control discrimination problems. The goal is to partition a given input graph in two sets, a clique and an independent set, such that there is no edge connecting a vertex of the clique with a vertex of the independent set. Following a parsimonious principle, we consider the problem that aims to modify the input graph into a most similar output graph that consists of a clique and an independent set (with no edge between the two sets). First, we present a theoretical result showing that the problem admits a polynomial-time approximation scheme. Then, motivated by the complexity of such an algorithm, we propose a genetic algorithm and we present an experimental analysis on simulated data.
Dondi, R., Mauri, G., Zoppis, I. (2016). Clique editing to support case versus control discrimination. In L.C. Jain, R.J. Howlett, I. Czarnowski, A.M. Caballero, L.C. Jain, L.C. Jain (a cura di), Intelligent Decision Technologies, Part I (pp. 27-36). Heidelberg : Springer Science and Business Media Deutschland GmbH [10.1007/978-3-319-39630-9_3].
Clique editing to support case versus control discrimination
Mauri, G;Zoppis, I
2016
Abstract
We present a graph-based approach to support case vs control discrimination problems. The goal is to partition a given input graph in two sets, a clique and an independent set, such that there is no edge connecting a vertex of the clique with a vertex of the independent set. Following a parsimonious principle, we consider the problem that aims to modify the input graph into a most similar output graph that consists of a clique and an independent set (with no edge between the two sets). First, we present a theoretical result showing that the problem admits a polynomial-time approximation scheme. Then, motivated by the complexity of such an algorithm, we propose a genetic algorithm and we present an experimental analysis on simulated data.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
