Seminars 2016

The next seminar will be held August 29, 2016. Please see below for details. See also this Google Calendar.

Welcome Seminar

Date/Time: Monday, August 29, 2016 from 4:10 p.m. - 5:00 p.m.

Location: Barnard Hall 108

Presenter: John Paxton, Gianforte School of Computing, Montana State University

Abstract: This seminar will provide new and continuing graduate students with (1) useful information, (2) an opportunity to meet other students, staff and faculty, and (3) an opportunity to ask questions.

Departmental Awards Seminar

Date/Time: April 25, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: John Paxton, Dept. Computer Science, MSU

Abstract: At the end of every academic year, we celebrate the accomplishments of our graduate students and faculty. Join us for this year's celebration where several departmental awards will be given. Refreshments will be served.

A New Discrete Particle Swarm Optimization Algorithm

Date/Time: April 18, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Shane Strasser, Dept. Computer Science, MSU

Abstract: Particle Swarm Optimization (PSO) has been shown to perform very well on a wide range of optimization problems. One of the drawbacks to PSO is that the base algorithm assumes continuous variables. In this paper, we present a version of PSO that is able to optimize over discrete variables. This new PSO algorithm, which we call Integer and Categorical PSO (ICPSO), incorporates ideas from Estimation of Distribution Algorithms (EDAs) in that particles represent probability distributions rather than solution values, and the PSO update modifies the probability distributions. In this paper, we describe our new algorithm and compare its performance against other discrete PSO algorithms.  In our experiments, we demonstrate that our algorithm outperforms comparable methods on both discrete benchmark functions and NK landscapes, a mathematical framework that generates tunable fitness landscapes for evaluating EAs.

Video recording available:

Using Machine Learning to Detect Distant Evolutionary Relationships between Protein Families

Date/Time: April 11, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Mensur Dlakic, Dept. Microbiology & Immunology, MSU

Abstract: Given the relative ease of genomic sequencing, improving the functional annotations of known proteins has become more important than the production of additional protein sequences. Since experimental determination of all protein functions is impractical, in most cases functional assignments are still done using only computational methods. We develop machine learning methods and software tools to find distant relationships between protein families at the level of sequence even in the absence of statistically significant scores. In addition, interactive web servers will provide general public with an easy access to these methods. The combination of human expertise and machine learning techniques will allow us in the long term to systematically catalog many of unclassified proteins and infer their biological functions.

Video recording available:

Discovery and Analysis of Communities in Evolving Political Contribution Networks

Date/Time: April 4, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Scott Wahl, Dept. Computer Science, MSU

Abstract: An important aspect of social networks is the discovery and partitioning of the complex graphs into dense sub-networks referred to as communities. The goal of such partitioning is to find groups who have similar attributes or behaviors. In the realm of politics, it is possible to group individuals with similar political behavior by analyzing campaign finance records. In this paper we show the effectiveness of hierarchical fuzzy spectral clustering over political contribution networks. The results show that clustering the data into two communities generally shows strong correlation between fuzzy membership values and existing estimates of political ideology. Further breakdowns of the data highlight different patterns of behavior. Analyzing these networks in time segments shows how individual behaviors and ideologies may change over time.

Agent-Based Modeling in Electrical Energy Markets Using Dynamic Bayesian Networks and Relevance Vector Machines

Date/Time: March 28, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Kaveh Dehghanpour, Dept. Electrical & Computer Engineering, MSU

Abstract: Electrical power Generation Companies (GenCos) compete with each other on wholesale electrical energy markets over the supply of electrical power to the consumers. Energy markets are based on auction mechanisms in which each player tries to maximize its immediate pay-off value by optimizing its bidding strategy. However, the complexity of the problem is due to the fact that the players need to make decisions under uncertain conditions: firstly, the players do not have access to their competitors' private cost data, which means that each player is a source of uncertainty to its competitors. Secondly, the actual value of electrical load is unknown, a priori, and needs to be predicted. In this project, a hierarchical agent-based framework is presented to model the decision-making problem of GenCos in an electrical energy market. Each GenCo is modeled as an agent with its private computational and decision making capabilities. The concept of Dynamic Bayesian Network (DBN) is employed to represent the "belief space" of GenCo agents. Each agent trains/updates its private belief system using relevance vector machine (i.e., sparse Bayesian learning). The trained DBN is then used to predict the best response to competitors for the incoming rounds of auction. It is shown that as the GenCo agents track their best response to their competitors the market approaches its Nash equilibrium over time.

Video recording available:

Hearing the Earth's Music Through Computing: Sonification of GPS Data

Date/Time: March 21, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Linda Antas, School of Music (Music Technology), MSU

Abstract: Sonification is the mapping of data onto musical/sonic parameters. It is used to assist in data interpretation where visual representations of data are problematic, insufficient, or unavailable. For the computer musician, sonification techniques provide diverse opportunities for research and creativity. The data used, mapping strategies, and how—if at all—to manipulate the data for best musical outcomes are all significant factors. Examples in this presentation will be drawn from a work in progress that sonifies GPS data collected in the Montana wilderness. The data is mapped using an algorithmic composition program that can output to a variety of formats, allowing it to be accessed as standard musical notation, or to be sent directly to a computer music programming language for creating synthesized sounds.

Video recording available:

Monothetic Clustering and Extensions to Clustering Functional Data

Date/Time: March 7, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Mark Greenwood, Dept. Mathematical Sciences, MSU

Abstract: Cluster analysis seeks to find groups of observations that are similar within and different among the created groups. Monothetic clustering algorithms create groups of observations that share common traits, in contrast to the more common polythetic clustering algorithms that create groups that are similar “on average”. The basic ideas and advantages of monothetic clustering are reviewed, including its connections to multivariate regression trees. Progress on the development of an R implementation will also be provided. High dimensional responses that can be considered as having been measured continuously over an argument, such as time or wavelength, are often called functional data. By dividing the functional domain into subregions and recursively partitioning the overall curves based on information from the subregions, a new algorithm, called Partitioning using Local Subregions (PULS), is developed. PULS seems to be competitive with other common functional clustering techniques and shares some characteristics with the monothetic clustering algorithm even though it is no longer monothetic. This is joint work with Tan Tran (Montana State University) and David Hitchcock (University of South Carolina).

The Science of Stories: The Narrative Policy Framework

Date/Time: February 29, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Elizabeth Shanahan (with Michael D. Jones), Dept. Political Science, MSU

Abstract: Narratives are the lifeblood of politics. However, not until the development of the Narrative Policy Framework (NPF) (co-developed by Drs. Shanahan-MSU, McBeth-ISU and Jones-OSU) have narratives been operationalized into a class of variables to test the impact of narratives on decision making. This presentation will provide an overview of the NPF, with two examples of research at the mico- and meso-levels of analysis. The former is an experimental design testing the effects of narrative strategies on individual opinion regarding the introduction of bison in the northern Montana prairie. The latter is preliminary network analyses of narrative constructions of 4 groups with different cultural cognitions (Kahan et al. 2011) regarding campaign finance reform.

Exploring Ethics in Data and Technology Research

Date/Time: February 22, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Sara Mannheimer, Scott Young, Jason Clark, Justin Shanks, University Libraries, MSU

Abstract: Technology is advancing at a pace so rapid that ethical inquiry is often left unaddressed. The MSU Library is currently conducting several research projects with user data, including Twitter sentiment analysis, personalized search development, and semantic web development. Over the course of these projects, ethical gray areas have emerged, prompting questions regarding ethical practice in “big data” research. Although huge amounts of data are freely available to us, much of it comes from human creators. Simply because we can use this data, does not necessarily mean that we should use it. Acknowledging this straightforward, but often overlooked dictum gives rise to various multifaceted ethical questions. This seminar will not only introduce ongoing data-oriented research occurring at MSU Library, but will also consider the broader ethical components embedded within research and product development. Who is affected by our research? Do users understand when and how their data is being used? How can we anticipate user expectations and values? What is the balance between personalized services and user privacy? Join us for a discussion of data-driven research and its ethical implications.

Secure Knowledge Management

Date/Time: February 8, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Dalal Alharthi

Abstract: Knowledge has become one of the most important driving forces for organizational success. Organizations are becoming more knowledge intensive. Therefore, taking care of knowledge is important for every organization nowadays. Knowledge must be managed. Knowledge management (KM) seeks to increase organizational capability to use knowledge as a source of competitive advantage. The challenge for organizations is to develop effective strategies for managing the knowledge.

Security of Information is a major concern for organizations nowadays. Secure knowledge management is one of the emerging areas in knowledge management and information system. It refers to the management of knowledge while adhering to principles of security and privacy. As KM has become a more central part of organizational activities, securing organizational knowledge has become one of the most important issues in the KM field. Knowledge needs to be protected so that it is properly secured from threats. Knowledge security addresses the protection of knowledge in organizations.

In this presentation, I will deal with knowledge management and knowledge security. I will try to explore different approaches used to secure knowledge management. Then, I will seek to identify a number of challenging security issues in Knowledge Management associated with protecting knowledge. Finally, I will illustrate the application of knowledge management and security by providing some examples from the national government of Saudi Arabia.

Establishing a Prospective, Long-term Follow-up, Pilot Study of Mental Health Biosignatures

Date/Time: January 25, 2016 from 4:10 p.m. - 5:00 p.m.

Location: EPS 108

Presenter: Matt Byerly, MD, Center for Mental Health Research and Recovery

Abstract: Mental illness affects 25% of the US population each year, 6% having serious mental illness.  It also strikes the young, with 50% developing illness by age 14 and 75% by age 24, making these illnesses the most disabling disorders before age 50, and most costly of all health conditions world-wide, with estimated annual US costs of more than $300 billion.  Montana has especially significant mental health challenges including the highest suicide rate in the country, at nearly twice the national rate; large populations at high risk of mental illness including Native Americans and military veterans, and; rural settings with limited mental health treatment resources.

The new MSU Center for Mental Health Research and Recover (CMHRR) is in the process of developing a prospective, long-term follow-up, pilot study of mental health biosignatures.  A biosignature, commonly used in multiple areas of medicine, is a unique combination of measureable, biologic features of a person and their illness that aids in making a diagnosis. To date, biosignatures are not used in the routine diagnosis and treatment of mental disorders.  This research will determine if we can match biologically-based diagnoses with response to specific treatments.  The result would be diagnostic “biosignatures” for individual patients that could identify their best initial treatment choice, speeding up recovery for each person with mental illness.  The work would also further identify brain signatures of illness development and progression that could aid in early and accurate diagnosis and, in turn, guide the focus of intensive preventive interventions for those at especially high risk.

Dr. Byerly, Director of the CMHRR, will discuss the background of biosignature work in mental illnesses, review the proposed study, and discuss potential relevance of the work to the computer sciences.


Seminars from 2015.