Page 42 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Science
Department of Physics

and to minimize the variance of the rating, i.e. maintain the fairness between different workers. For the simple version
of the problem, which the different players have the same preferences and different days of a week have the same
preferences by the players, it is found that there are 29064960 optimal results, using exhaustive search. For the more
realistic version of the problem, genetic algorithm within 5000 generations is used, with compensation calculated in the
fitness if constraints of scheduling are violated. It is found that mutation method is better than crossover in this problem.

Supervisor: SZETO Kwok Yip / PHYS
Student: WANG Ziqi / PHYS Course: UROP1100, Summer

In this paper we will discuss a network model of a financial system problem, study the construction of the 3 classical
types of network: Erdo˝s–R´enyi(ER) model, Watts-Strogatz (WS) model and Baraba´si–Albert (BA) model and focus on
the relation between structure and the model stability of the networks in the financial problem. Then, we will apply the
genetic algorithm on the problem and figure out the factor that affects the stability of the system. The cascade failure
phenomenon is also a potential risk we need to be studied and avoided in the models. We aim to find the efficient
rewiring strategy to reduce the cascade failure through programming, such as genetic algorithm and enumeration.

Supervisor: SZETO Kwok Yip / PHYS
Student: WONG King Chun / PHYS Course: UROP1100, Spring
UROP2100, Summer

Opinion formations in a social networks, such as elections, are highly non-equilibrium processes that exhibit rich
dynamics and many interesting features. Borrowing the ideas and techniques of the Ising model from statistical
mechanics, a model was built to study the voting dynamics between two competing parties. Based on this model, we
had studied the effect when one of the agents in the network is fixed to belong to a certain party immutably, that is, the
effect of the implantation of a “spy”. A party-specific centrality measure was introduced to analyze the effect of different
implantation sites and the affected dynamics. We ended this report with a note of possible future developments of the
project.

Supervisor: SZETO Kwok Yip / PHYS
Student: XIA Yusen / SSCI Course: UROP1000, Summer

This report is a summary of the work done in UROP 1000 course, summer 2018. We preliminarily study the Evolutionary
Genetic Algorithms. We study the conventional GA which based on mutation and crossover, and the adaptive Mutation
Only GA introduced by Prof Szeto Kwok Yip. In both methods, we establish a set of chromosomes to represent the
solutions and try to keep enough information of the fittest as well as introduce new information to the population by
mutation. A probability matrix is introduced in the latter method and it is parameter free and evolving with time. We
apply the conventional method to a function optimization problem and both the conventional and Prof. Szeto’s methods
to Travelling Salesman Problem. Different improvements of efficiency are discussed in the TSP experiments.

Supervisor: SZETO Kwok Yip / PHYS
Student: ZHANG Tianpeng / COSC Course: UROP3100, Fall
UROP4100, Summer

The contention resolution problems deal with sharing limited resources among multiple parties. Some cases can be
solved using centralized methods, or using communication between parties, but there are problems that are inherently

32