NETWORK

APPLICATIONS
PROJECT
«Petri Nets Based
Max-flow/Min-cut Modeling and Analyzing»

Presentation by Eylül Karahan

STD Number: 192010050013
 CONTENTS
• Abstract • Simulation and Analysis
• Introduction • Validation
• Problem Statement and • Conclusion and Future
Preliminaries Work
• The Proposed Modeling Method
 ABSTRACT
This paper introduces a Petri Nets (PNs)-based method for modeling and analyzing the
max-flow/min-cut problem in distributed systems. The method involves generating
models from a flow network and its residual network, combining them to simulate the
classic Ford-Fulkerson for solving the problem. The method is presented as an intuitive
and effective way to understand and present the max-flow/min-cut theory, and is
extensible for other graph problems.
 INTRODUCTION
The maxflow/min-cut problem in graph theory is a widely explored combinatorial optimization problem. It has been applied in various
scenarios, such as road networks evaluation, online semisupervised learning, invulnerability analysis of power grids, and energy
minimization in vision. The concept of employing min-cut/max-flow to minimize certain energy functions was first introduced by Greig et
al. for image restoration, and later by Boykov et al. for image segmentation, stereo with occlusions, and multi-camera scene reconstruction.

This paper aims to use Petri Nets (PNs) as a solution to formalize the max-flow/min-cut problem and depict its solutions. PNs are a
mathematical and graphical modeling language with powerful analysis techniques, making them intuitive for describing abstract knowledge
and modeling procedures. Simulations of PNs can be defined mathematically, making the analysis more convincible.

The paper introduces PNs models for flow networks and residual networks, which are combined to simulate the Ford-Fulkerson method for
solving the max-flow/min-cut and finding different flow distributions when the max-flow achieves. A back-flow strategy is also presented to
identify different flow distributions when the max-flow achieves, and formal analyses and proofs are made to validate the proposed method.
INTRODUCTION
 PROBLEM STATEMENT AND PRELIMINARIES
This section discusses the max-flow/min-cut problem, a mathematical model used to describe
distributed systems. The max-flow problem focuses on the flow in a network, with the goal of
finding a flow of maximum value. The method is closely related to the Ford-Fulkerson method,
which relies on residual networks. The residual network of a flow network is induced by a flow,
with the flow and capacity being represented by f and c, respectively. The goal is to determine the
largest number of crates per day that can be shipped and produced.

Petri Nets, introduced by Dr. Carl Adam Petri in 1939, is a mathematical modeling tool for
describing distributed systems. It consists of basic elements such as Places and Transitions, and
tokens can be transmitted between them through transitions.
PROBLEM STATEMENT AND
PRELIMINARIES
 PROBLEM STATEMENT AND PRELIMINARIES

Ford Fulkerson Method:

The Ford-Fulkerson method, proposed by L.R. Ford and D.R. Fulkerson in 1962, is a classic solution for max-flow/min-cut. It acquires
the maximum flow in a flow network by finding an augmenting path in the residual network and augmenting it iteratively. The method is
based on the Max-Flow/Min-Cut theorem, which states that if f is a flow in a flow network, it must satisfy certain conditions.
 THE PROPOSED MODELING METHOD
In this Section, PNs based modeling methods for (1) flow network and (2) its corresponding residual
network are proposed.
 THE PROPOSED MODELING METHOD
In this Section, PNs based modeling methods for (1) flow network and (2) its corresponding residual
network are proposed.
 SIMULATION AND ANALYSIS
The method focuses on finding a max-flow/min-cut solution in flow networks using PNs-based modeling systems.
It demonstrates this by illustrating a solution intuitively through simulation and providing analyses. The method
initializes the source Place Ps with enough tokens and allows the system to run under PNs rules until there is no
path for a token flow from source Place Ps to sink Place Pt. A max-flow occurs when there is no path for a token
flow from source Place Ps to sink Place Pt. A min-cut is found by dividing V into two categories, CS and CT. The
method also acquires a flow value of each edge when maximum flow achieves, using a back-flow strategy.
Different back-flow paths may lead to different flow distributions. An integrated system for max-flow/min-cut is
proposed using the back-flow strategy. The top part of the system contains a flow network model and the bottom
part contains a residual network model. Places and transitions are added. The system runs in two phases if the
source Place is initialized with enough resources. In the first phase, tokens flow in the residual network NGf, and
in the second phase, tokens flow in the flow network NG. The system triggers a reset of the flow.
SIMULATION AND ANALYSIS
SIMULATION AND ANALYSIS
 VALIDATION
The method aims to find a max-flow/min-cut solution in flow networks using PNs-based modeling systems. It
initializes the source Place Ps with enough tokens and runs under PNs rules until no token flow is found. A max-
flow occurs when no path is found, and a min-cut is found by dividing V into CS and CT categories. An integrated
system is proposed using a back-flow strategy, running in two phases.
Three theorems were used to validate the method used in this study and the accuracy of all three theorems was
proven.
 CONCLUSION AND FUTURE WORK

This paper proposes a PNs-based approach for modeling and analyzing the max-flow/min-cut
problem, a basic graph theory used in various research. The paper aims to formalize the problem
and depict its solution, using animated graphical notations and mathematically defined simulations.
Novel modeling methods are introduced for flow networks and residual networks, and PNs models
are combined to simulate a classic max-flow/min-cut solution. The results show that a max-flow
achieves when there is no path for a token transmit from source Place to sink Place, a min-cut can
be determined by liveness analysis of the system, and different flow distributions depend on the
choice of different back-flow paths. Proofs of the simulation results validate the feasibility of the
proposed method. The paper aims to extend this work for other graph theory problems and apply
PNs for state-of-the-art intelligent algorithms.
THANK YOU