Current student research projects
Sam DeFabbia-Kane (Research project, Fall 2009; advisors: Norman
Danner, Danny Krizanc)
Tor (The Onion Router) is an anonymity network. It protects the
anonymity of its users by routing their network connections through
a circuit of three proxies located around the world. There have been
many attacks proposed against Tor aimed at compromising the anonymity
it provides. One of the most common types of attack is an end-to-end
correlation attack. If an attacker controls the first and last nodes in
a circuit, they can compare the network streams at each and identify the
sender and the recipient of the stream, compromising its anonymity. One
way for an attacker to increase the chance of that happening is for them
to kill any circuits they have nodes present in but can't control. The
user is then required to create a new circuit, giving the attacker another
chance to compromise it. We're working on a method of detecting attackers
doing that that's feasible to implement and run on the Tor network.
Carlo Francisco (Research project, Fall 2009; advisor: Danny Krizanc)
A central question in microbiology is finding a consistently reliable method
to demarcate the groups playing ecologically distinct roles (ecotypes) in a
set of bacterial cells within a natural community. Multiple programs have
been developed for this purpose. We want to test the accuracy of four of
these programs, Ecotype Simulation, AdaptML, GMYC, and BAPS, on real and
simulated data. The goal is to see which one yields closer results to the
actual known demarcations of the clades that will be tested. Since bacterial
cells can be isolated from the habitat they're specialized to, we make use
of a variable called "specialization," or the probability that a particular
kind of cell (differentiated by its DNA sequence) is isolated in this way.
We also intend to include the relative running time of each of the
algorithms in the study.
Ted Nichols
Image Processing for Agent Navigation
(Research Project, Fall 2009; advisor: Eric Aaron)
When an intelligent agent is navigating a virtual environment, it
has access to whatever it needs to know about its own location and
the location of potential obstacles. An agent navigating in the real
world, however, has to get the same information based solely on the
input it receives from its sensors. Visual feeds can provide a great
deal of information to an agent, provided that the raw image data can
be transformed into a useful model of the agent's environment. We are
working on implementing computer vision techniques to allow an agent to
gather the information it needs to successfully navigate its environment.
Juan Pablo Mendoza, Walls and Polyhedral Obstacles in Dynamical
Autonomous Navigation
(Hughes Summer Program 2009; Research Project, Fall 2009;
advisor: Eric Aaron)
Previous research has explored the problem of simulating autonomous
navigation of agents with sub-deliberative intelligence. These agents
do not plan their paths, but instead they only react to their
environment. The dynamical systems approach to autonomous navigation
is a model of reactive navigation that has desirable characteristics
such as dynamical decision-making and resistance to the local minima
problem. It has an implicit assumption, however, that the environment
is composed only of circular obstacles. Our work introduces a
time-efficient method to incorporate walls and polyhedral obstacles to
the dynamical systems approach to navigation with a high
target-reaching rate. Agents navigating using this method perform
better than agents using previously employed circular approximations
for non-circular obstacles.
Past student research projects
Henny Admoni, Demonstrations of Dynamical Intention for Hybrid Agents
(M.A. thesis, 2009; advisor: Eric Aaron)
Representations of intention shared by reactive and deliberative systems
of hybrid agents enable seamless integration of high-level logical
reasoning and low-level behavioral response. This thesis presents an
architecture for hybrid dynamical cognitive agents (HDCAs), hybrid
reactive/deliberative agents with cognitive systems of continuously
evolving beliefs, desires, and intentions based on BDI and spreading
activation network models. Dynamical intentions support goal-directed
behavior in both reactive and deliberative systems of HDCAs: on the
reactive level, dynamical intentions allow for continuous cognitive
evolution and real-time task re-sequencing; on the deliberative level,
dynamical intentions enable logical reasoning and plan generation. Because
intention representations are shared between both systems, reactive
behavior and goal-directed deliberation are straightforwardly integrated
in HDCAs. Additionally, Hebbian learning on connections in the spreading
activation network of beliefs, desires, and intentions trains HDCAs’
reactive systems to respond to typically deliberative-level information.
To establish comparability between HDCAs and traditional BDI-based
architectures, dynamical intentions are shown to be consistent with the
philosophical definition of intention from other BDI models. Simulations
of autonomous, embodied HDCAs navigating to complete tasks in a grid city
environment illustrate dynamical, intention-based
behavior that derives from clean integration of reactive and deliberative
systems.
Adam Robbins-Pianka, Modeling mRNA Secondary Structure Effects on Translation
Initiation in Saccharomyces cerevisiae
(M.A. thesis, 2009; advisor: Mike Rice)
This M.A. thesis studies the influence of RNA secondary structure on
translational events in Saccharoymces cerevisiae.
Jon Gillick,
A Clustering Algorithm for Recombinant Jazz Improvisations
(Honors thesis, 2009; advisor: Danny Krizanc)
Music, one of the most structurally analyzed forms of human creativity,
provides an opportune platform for computer simulation of human artistic
choice. This thesis addresses the question of how well a computer model
can capture and imitate the improvisational style of a jazz soloist.
How closely can improvisational style be approximated by a set of rules?
Can a computer program write music that, even to the trained ear, is
indistinguishable from a piece improvised by a well-known player?
We discuss computer models for jazz improvisation and introduce a new system,
Recombinant Improvisations for Jazz Riffs (Riff Jr.), based on Hidden Markov
Models, the global structure of jazz solos, and a clustering algorithm.
Our method represents improvements largely because of attention paid to
the full structure of improvisations.
To verify the effectiveness of our program, we tested whether listeners
could tell the difference between human solos and computer improvisations.
In a survey asking subjects to identify which of four solos were by Charlie
Parker and which were by Riff Jr., only 45 percent of answers among 120
people were correct, and less than 5 percent identified all four correctly.
Henny Admoni, Decision Making and Learning for Hybrid Dynamical Agents (Hughes Summer Program 2007; Honors Thesis, General Scholarship, 2008; advisor: Eric Aaron)
Models of human decision making and learning may benefit from being implemented in cognitively-inspired, dynamic frameworks. This paper describes one such framework built on a hybrid dynamical system, which models agent cognition as the continuous evolution of cognitive elements with occasional discrete changes in patterns of behavior. Cognitive elements are beliefs, desires, and intentions, from the Belief-Desire-Intention theory of human rationality, along with ground concepts and sequencing intentions. Continuous dynamics of the hybrid system result from changes in element activation levels; activation can spread to related elements over links in a neurologically-inspired spreading activation network. This hybrid system performs decision making by selecting new patterns of behavior when activation levels of elements reach certain configurations, and performs learning by strengthening spreading activation links between elements. This system is capable of representing long-term sequences of actions, or plans, as well as dynamically replanning by selecting new actions when environmental changes make old action sequences undesirable. In simulations of the model, agents navigated grid world environments while dynamically selecting goals and actions to fulfill those goals, and autonomously learned associations between people and places in their environments.
Bach Dao, A Simulation Study of Protein Family Degree Distribution
(Honors thesis, 2008; advisor: Danny Krizanc)
The focus of this thesis is the degree distribution of protein family
network graphs. We propose three models of evolution that generate a
protein family. The first one uses Sequence Alignment to quantify protein
relationship while the second uses the number of mutations accumulated on
proteins. The last model incorporates explicitly preferential attachment.
Following many other studies, we consider three operations: duplication,
gene death and mutation. The main result that we report from the three models
is that although the exponential distribution is the best fit of the data,
the power law distribution fits the data well with the rates of evolution
found in many studies.
Jesse Farnham,
Performance Enhancement and Equivalence Criteria for Cellular
Automaton-Based Tumor Simulations
(Hughes Summer Program 2007; Honors thesis, 2008; advisor: Eric
Aaron)
This paper describes a performance enhancement to a cellular
automaton-based simulation of tumor growth. A cellular automaton is a
grid-based data structure whose elements are updated according to specific
rules. Elements represent small areas of simulated tissue and contain
local cell population and nutrient concentration data for those areas. The
simulation's computational efficiency is enhanced by suppressing updates
to \emph{steady-state} tissue locations, areas with little change in
nutrient concentration or cell population. This modification produces
different tissue development from the original, canonical method in which
all tissue areas are updated on every timestep. This paper defines several
criteria by which a modified, efficient simulation can be considered
equivalent to an original, canonical method; these criteria are applied
to determine whether or not the performance enhancement can be considered
equivalent to the original simulation.
Daniel Hore, Implementation and Type-checking for ATR
(Honors thesis, 2007; advisor: Norman Danner)
In this thesis we implement an interpreter and describe a principal-type
algorithm for the language ATR (for Affine Tiered Recursion)
defined by Danner and Royer. The interpreter and the principal type
algorithm (when implemented) will allow programmers to program in ATR .
This becomes interesting upon understanding the constraints guaranteed for
any program that successfully compiles in ATR, namely that any program
that compiles in ATR runs in polynomial time. However, the notion of
polynomial time must be extended since ATR contains higher order types.
Brendan Dolan-Gavitt, Timing Attacks in Anonymity-Providing Systems
(Honors
thesis, 2006; advisors: Norman Danner, Danny Krizanc)
In general, when a client accesses any server over the Internet, a
great deal of information is trivially obtainable. In particular,
the sender's IP address is included in every packet of data they
transmit, and this information can be used to determine their
Internet Service Provider and in many cases their approximate
geographic location. In addition, each packet contains the IP
address of its destination. Anonymous networks provide a way of
preventing outside observers from determining that an initiator and
a responder are, in fact, communicating.
The most common way of achieving this anonymity is by routing the
connection through a series of intermediate proxy servers,
along the way mixing or delaying
packets in order to frustrate attempts at uncovering the
sender through traffic analysis.
However,
for most interactive applications, such as web
browsing, SSH sessions, or live chat, too much mixing or delaying
will result in a poor or unusable experience for the user. A web
browsing session, would quickly become frustrating if it took five
minutes for each page to load. It seems, then, that there is a
trade-off between anonymity and speed.
This thesis examines what threats are
actually posed by traffic analysis against low-latency mix-nets and
how practical it is to implement them on a real network, using Tor
as our case study.
Adam Robbins-Pianka and Paul Cao, Analysis of upAUG Conservation and
Possible Misannotation of S. cerevisiae ORFs (Hughes Summer Program 2006;
advisors: Mike Rice, Michael Weir)
In order to study misannotation, we first identified candidate genes by
comparing alignments of orthologous genes of four Saccharomyces species
and finding well-conserved sequences which contained possible start sites
upstream of the annotated open reading frame (annORF). Using information
theoretic analysis, we compared and contrasted patterns in the sequence
data surrounding the upAUGs and their corresponding annotated AUGs
(annAUGs) of the candidates with those of high-confidence, verified
genes to further refine our assessment of misannotation. Furthermore,
we studied the sequential pattern of conservation of upAUGs in order
to discover possible introns. To further identify and confirm intron
candidates, we aligned upstream protein sequences of orthologous genes
to identify regions of extended conservation and, conversely, gaps of
non-conservation, which may identify regions of introns. It is our aim
in this project to improve upon current gene and intron annotation by
identifying and studying the characteristics of upAUGs in the yeast
genomes.
Other student projects
The Multitouch Project
The Wesleyan Multitouch Developers group is exploring the future of
human-computer interaction through natural user interfaces. As the
hardware for creating touch-sensitive surfaces becomes cheaper and more
popular, more possibilities for intuitive interactions with computers
emerge. We are currently working on a framework for well-integrated,
efficient, and usable software.
The Humanitarian FOSS project
We are part of a growing community involved in The Humanitarian FOSS
Project, dedicated to building and using Free and Open Source Software
(FOSS) to benefit humanity. We are a founding chapter of the H-FOSS
Project, along with Trinity College and Connecticut College. See
hfoss.wesleyan.edu for more
information.
|
![[Wesleyan University]](http://www.wesleyan.edu/templates/utilities/navbar_2009/logo.gif "Wesleyan University front door"){kind=logo}
