# TU Darmstadt ULB TUbiblio

# Partitioning into Isomorphic or Connected Subgraphs

## Abstract

This thesis deals mainly with the partitioning and connectedness of graphs. First, we show that the problem of partitioning the nodes of a graph into a specific number of subsets such that the induced subgraphs on these sets are isomorphic to one another is NP-complete. If the induced subgraphs have to be connected, the problem remains NP-complete. Then we inspect some special graph classes for which the problem is solvable in polynomial time. Afterwards, we deal with the problem of defining a polytope by incidence vectors of nodes, which induce a connected graph. We inspect some facet-defining inequalities and their general structure. For some graph classes we state the full description. We then proceed to the problem of partitioning the nodes of a graph into a given number of parts such that the induced graphs are connected. For the corresponding polytope we show the dimension and some facet defining inequalities. This theoretical inspection is advanced by the problem of partitioning a graph into different parts such that the parts induce a connected graph in order to maximize the induced cut. We introduce different ideas for solving this problem in SCIP and show the numerical results. This leads to interesting problems on MIPs in general. As the problem in literature generally deals with the feasible region, we focus on the objective function. To do that, we inspect the problem of finding MIPs for problems with nonlinear objective functions. We discuss properties and requirements showing the existence or non-existence of particular formulations. Lastly, we inspect the problem of partitioning the nodes of a graph such that all but one class are isomorphic. This problem becomes interesting if the part not inducing the isomorphism is minimized. For this purpose we also introduce a technique, which generates the parts by brute-force. Instead of partitioning the graph into isomorphic parts, we proceed to the problem of similar graphs. In this case we inspect different similarities and show algorithms which implement these.

Item Type: Ph.D. Thesis
Erschienen: 2019
Creators: Lüthen, Hendrik
Title: Partitioning into Isomorphic or Connected Subgraphs
Language: English
Abstract:

This thesis deals mainly with the partitioning and connectedness of graphs. First, we show that the problem of partitioning the nodes of a graph into a specific number of subsets such that the induced subgraphs on these sets are isomorphic to one another is NP-complete. If the induced subgraphs have to be connected, the problem remains NP-complete. Then we inspect some special graph classes for which the problem is solvable in polynomial time. Afterwards, we deal with the problem of defining a polytope by incidence vectors of nodes, which induce a connected graph. We inspect some facet-defining inequalities and their general structure. For some graph classes we state the full description. We then proceed to the problem of partitioning the nodes of a graph into a given number of parts such that the induced graphs are connected. For the corresponding polytope we show the dimension and some facet defining inequalities. This theoretical inspection is advanced by the problem of partitioning a graph into different parts such that the parts induce a connected graph in order to maximize the induced cut. We introduce different ideas for solving this problem in SCIP and show the numerical results. This leads to interesting problems on MIPs in general. As the problem in literature generally deals with the feasible region, we focus on the objective function. To do that, we inspect the problem of finding MIPs for problems with nonlinear objective functions. We discuss properties and requirements showing the existence or non-existence of particular formulations. Lastly, we inspect the problem of partitioning the nodes of a graph such that all but one class are isomorphic. This problem becomes interesting if the part not inducing the isomorphism is minimized. For this purpose we also introduce a technique, which generates the parts by brute-force. Instead of partitioning the graph into isomorphic parts, we proceed to the problem of similar graphs. In this case we inspect different similarities and show algorithms which implement these.

Divisions: 04 Department of Mathematics
04 Department of Mathematics > Optimization
04 Department of Mathematics > Optimization > Discrete Optimization
Date Deposited: 03 Mar 2019 20:55 View Item