Fall 2012

Tuesday, August 28th - Math & CS Welcome Back BBQ
Join us at Noon at the KW Lawn for a BBQ.  The Math & CS professors will grill hot dogs, hamburgers & veggie burgers as well as supply side dishes & desserts.  Everyone welcome! Come out & join the fun!

Tuesday, September 4th —"Summer REU's"
Natalie Stanley will present "The Role of Incoherent Micro RNA Feed --- forward Loops in Gene Regulatory Network Stability"

Abstract:  Incoherent micro RNA feed--‐forward loops (miRNA FFLs) are a recurring motif in gene regulatory networks that link a transcription factor, a micro--‐RNA and a target gene. These FFLs have been shown to contribute to network stability. Specifically, miRNA FFLs assist in the maintenance of optimal protein levels in the cellular environment of an organism by providing a buffer against extrinsic and intrinsic noise. The stability of a gene network is correlated with a relatively larger number of initial conditions leading to a steady state or a limit cycle (basins of attraction). It can be shown that the addition of miRNA FFLs also increases the maximum basin of attraction size and therefore indicates stability. In this study, we examined how miRNA FFLs contributed to gene regulatory network stability from a theoretical standpoint. Finally, a numerical simulation was performed on the sensory organ precursor network in drosophila to justify the stabilizing effect of miR--‐7 in a biologically feasible context.

AND

Yujia Zhou will present "Quality Control of Sensor Data and Data Provenance Tracking"

Abstract: Scientists often rely on sensors to obtain data. Sometimes, sensors may go wrong, thus the raw data needs to be processed before it can be used. In this research, we used the 15-minute real-time data from the meteorological station at Harvard Forest and studied the quality control methods, including calibration adjustment, detection of irregular values, and gap filling of missing data. We developed R programs to detect and fix quality control problems. This process will be performed multiple times in the future due to improvements of quality control techniques and hence will generate different versions of datasets. However, as the datasets grow larger and time passes, it becomes difficult to know how a particular version of the dataset was derived from the raw data. As a result, recording the data provenance is necessary for scientists to understand data derivation. Data Derivation Graphs (DDG) can record the full provenance of how each data point is derived from the raw data, allowing scientists to keep track of their data. To accomplish this goal, we built a process simulating scientists’ initial processing of the raw data and the reprocessing after some of the quality control techniques are updated. We implemented this process in both Kepler and Little-JIL to compare the data provenance graphs they produce from identical processes. We found that while Kepler is easier to use for scientists with no programming background, Little-JIL has a much stronger visualization tool for drawing comprehensive DDGs and stores more information in them. 

12:00 p.m. to 1:00 p.m.
Tome 115
Lunch provided

Thursday, September 20th
Dick Forrester, Associate Professor of Mathematics, Dickinson College and
Amity Fox, Assistant Director/Intership Coordinator of Dickinson College Career Center
"Where Do I Go From Here?"

Abstract: In this chat they will discuss a wide variety of careers and opportunities for students majoring in mathematics and computer science.  In addition, they will talk about graduate school options, internships, and REUs (Research Experience for Undergraduates).  Specific information about our recent graduates will be provided.

12:00 p.m. to 1:00 p.m.
Rector Lecture Room (Stuart 1104)
Lunch provided

Tuesday, September 25th
Justine Heritage '14
"Graph 500 Performance on a Distributed-Memory Cluster"

Abstract:  Efficient data access is important for many modern applications. A network analysis of Facebook, for example, might rely on a quick search of a large graph structure with roughly 1 billion vertices representing users, and edges representing each of their friendships. The Graph 500 benchmark evaluates a computer's performance on problems such as these by measuring the number of traversed edges per second (TEPS) of large random graphs. We examine the implementation and execution of this benchmark on the distributed-memory cluster tara at the University of Maryland, Baltimore County's High Performance Computing Facility.

AND

Danfei Xu '14
"Object Recognition by Touch Using Bayesian Exploration"

Abstract: In order to endow robots with human-like tactile sensory abilities, they must be provided with tactile sensors and intelligent algorithms to select and control useful exploratory movements and interpret data from all available sensors. Current robotic systems do not possess such sensors or algorithms. In this study we integrate multimodal tactile sensing (force, vibration and temperature) from the BioTac® with a Shadow Dexterous Hand and program the robot to make exploratory movements similar to those humans make when identifying objects by their compliance, texture, and thermal properties. A statistical learning algorithm called Bayesian exploration is used to intellectually select movements that provide the most disambiguation between likely candidates of objects. The robot correctly identified 10 different objects on 99 out of 100 presentations.

12:00 p.m. to 1:00 p.m.
Tome 115
Lunch provided

Tuesday, October 9th
Christian Millichap '08
"What is a Manifold?"

Abstract: Manifolds are spaces that locally look like Eucliden space.  Such objects arise in mathematics in a variety of forms: curves, the surface of a doughnut, the solution space of some set of conditions, and even the space surrounding a knot.  In this chat, we will examine different ways to construct 1-dimensional and 2-dimensional manifolds.  We will also consider how to distinguish one manifold from another.  Afterwards, we will jump into the strange world of 3-dimensional manifolds.  3-dimensional manifolds are often difficult to visualize, but sometimes can be understood by imagining what it is like to live inside of one.  Throughout this chat, looking at examples and using visual intuition will be our main tools in understanding manifolds.

12:00 p.m. to 1:00 p.m.
Tome 115
Lunch provided

Thursday, October 18th (Stafford Lecture Room - Stuart 1104)
Jing Hu, Franklin and Marshall
"Predicting the Effects of Frameshifting Indels Using A Machine Learning Approach"

Abstract: Small insertions/deletions (indels 20 bp or less) account for nearly 24% of known Mendelian disease mutations. It is the second largest class of mutation type that leads to disease, following amino acid substitutions which account for over half of known Mendelian disease mutations. There exist many bioinformatics algorithms that predict whether an amino acid substitution affects protein function, and these are commonly used for predicting and prioritizing disease variants, but very little work has been done for indels. 

Each human has ~50-280 frameshifting indels, yet their implications are unknown. With inexpensive and ubiquitous genome sequencing, it would be time-consuming to analyze these hundreds of mutations manually, yet it would be important to distinguish the functionally neutral indels from those that are under negative selection.

In this talk, I will present SIFT indel, a Decision Tree based prediction method for frameshifting indels which has 84% accuracy, 90% sensitivity and 81% precision. We applied the SIFT Indel algorithm to the frameshifting indels identified from the human genomes sequenced by the 1000 Genomes Project (1000G) and by Complete Genomics (CGI). The results show that the percentage of FS indels predicted to be gene-damaging is negatively correlated with allele frequency. We also show that although the first frameshifting indel in a gene causes loss of function, there is a tendency for the second frameshifting indel to compensate and restore protein function.

12:00 p.m. to 1:00 p.m.
Stafford Lecture Room - Stuart 1104
Lunch provided

Wednesday, October 31st  
Peter W. Bates, Michigan State University
"Seeking to Understand the Processes of Life through Mathematics"

Abstract: For several hundred years, mathematical models have been used to seek understanding of the physical world.  Physical laws are postulated, translated into equations, and the equations solved or at least used to predict outcomes, which were then tested against observations.  This method of inquiry made its way into biological modeling, most notably in ecology and infectious diseases where mechanisms are at a macro-level and are fairly well-established.  Here we discuss biological processes as a cellular and molecular level, focusing on the modeling and simulation of some of those processes.

1:30 p.m. to 2:30 p.m.
Tome 115
Snacks provided

Tuesday, November 13th
Chandra Erdman
Talk Title: "Predicting Response Propensities and Setting Response Rate Expectations in Large National Surveys"

Abstract: The U.S. Census Bureau’s national surveys include the American Community Survey, the Current Population Survey, the Consumer Expenditure Survey, the Survey of Income and Program Participation, the National Health Interview Survey, and the National Crime and Victimization Survey. It is evident from their names that these surveys are diverse in content and level of privacy, and this leads to variation in response rates. Given increasing cost and other resource constraints, there is a clear need to have a good understanding of expected response rate variation that may affect decisions on data collection efforts.  Understanding this variation can be useful in determining when additional field effort on a case may not be cost-effective and in setting realistic interviewer performance expectations.  This talk presents the statistical techniques used by the U.S. Census Bureau in predicting survey case response propensities and in classifying small areas into strata to establish performance expectations for interviewers.

12:00 p.m. to 1:00 p.m.
Tome 115
Lunch provided

Thursday, November 29th (Stafford Lecture Room - Stuart 1104)
Dr. Greg Wilder
"Computational Creativity: Teaching Computers to Listen"

Abstract: Dr. Wilder will discuss factors that make musical data mining unique and describe an autonomous system that overcomes present challenges through the application of research in music cognition as a ground plan for intelligently adaptive algorithms. His presentation will also relate his experiences bringing these innovations to the music industry and layout a vision for future exploration in the field of computational creativity.

12:00 p.m. to 1:00 p.m.
Stafford Lecture Room - Stuart 1104
Lunch provided