Page 43 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Science
Department of Physics

decentralized and there is no efficient communication mechanism between different parties. For example to passengers
deciding among themselves who to take the only empty seat on the subway, or web clients attempting for the access to
the database. A common solution is to let every party follows the same randomized protocol, which is to toss a coin to
decide whether to acquire or yield the resource. The problem is resolved when only one party decides to acquire and all
others yield. The randomized protocol is extremely efficient when the size of contention, i.e. number of parties
attempting the same resource, is known. However, it is often the case that the contention size is hardly known by
individual parties, as in the contention of web clients, and in this work we try to tackle this issue. We will first review the
contention resolution problem, analyze the traditional randomized protocol, then introduce new two protocols which
has good efficiency without the knowledge of contention size. The idea is to adapt the probability of acquiring the
resource and the contention size through iterations, so that they match each other.

Time Dependent Mean Variance Analysis in Econophysics

Supervisor: SZETO Kwok Yip / PHYS
Student: WANG Juntao / PHYS Course: UROP1100, Spring

The portfolio management problem is a hot and complex topic in asset management and investment. In this paper, we
introduce a dynamic set of portfolio strategy for two stocks and cash using Genetic Algorithm. The two stocks are
selected from Hang Seng Index constituents, which satisfy two conditions: they are highly anti correlated during some
periods of time, and there are enough daily close prices data to be studied. The strategy comprises of three parts:
conditions, fraction, and action. The genetic algorithm is applied to search for optimal strategies, by which one can
achieve the largest Sharpe ratio. By employing this dynamic set of strategies, we show that the return rate of
investments is increased, compared to using random strategies.

Chaotic Dynamics in a Complex Network
Supervisor: SZETO Kwok Yip / PHYS
Student: WONG King Chun / PHYS Course: UROP4100, Fall

A model of analyzing quasi-periodic time series was constructed based on Parrondo’s paradox. The model consists of
two independent predictors, a fair guesser A and a history-dependent B. The goal of the model is to investigate the
possibility of Parrondo Effect in time series prediction, in which two losing prediction strategies can be combined to a
winning one, and to utilize the phenomenon to search for a optimal prediction scheme given a limitedly accessible
history. Simulations and analysis of the model had been performed by inputting artificial sinusoidal time series, and an
optimal mixture of the two predictor, A and B, had been attempted to construct both stochastically and deterministically.

Negative Refractive Index Meta-materials

Supervisor: TAM Wing Yim / PHYS
Student: BRACKE Vincenz / PHYS Course: UROP2100, Fall
UROP3100, Spring
UROP4100, Summer

In this progress report, I will briefly introduce the project I have worked on during this summer semester. I will introduce
the aim of the project and outline how what to achieve this. Then I will summarize the progress made this summer in
manufacturing a larger and thicker sample of a negative refractive index material using the combined focused ion beam
and scanning electron microscope system and detail the results gathered so far. Using pictures taken during the

33