Search Google Appliance


Seminar Archive 2008

 

 


Date: December 5, 2008

Location: Harder House, Room 104

Time: 12-12:50

Presenter: Robin Fenske

Title: Systems Science Community

Abstract: The general topic is: How communication platforms help us strengthen our Systems Science community AND How to communicate with non Systems Scientists about this crazy grad program you're involved in.

It's hard to communicate within our own Systems Science community about what we are all working on, and to express in a cohesive way the multitude of ideas from all the classes we've taken. And it's even harder to talk to our relatives and friends about what we are interested in. As espoused generalists, why is this so hard to do? What communication platforms might help us strengthen our own Systems Science community? Let's talk about it!

Bio: Robin Fenske is a 2nd year PhD Student, interested in understanding consumer behavior (and the level of conscious decision-making), intention, and assumptions, and how this understanding can help promote sustainable food systems and lessen rampant consumerism.

File attached or Link to Recording?
None

 



Date:
November 21, 2008

Location:
Harder House, Room 104

Time:
12-12:50

Presenter(s):
Louis Macovsky, DVM, MS

Title:
From Computer Model to Laboratory Model: Action at a Distance in Metapopulation Theory

Abstract:
Built upon metapopulation theory, a toxicant-dosed model was created to explore the range of possible dynamics of populations in sites contaminated by chemical toxins. A laboratory model using the beetle Tribolium castaneum was then created to partially test simulation results. Both the computer and the laboratory models support the “action at a distance” hypothesis, which states that mortality in one subpopulation has ecologically significant effects on nondosed subpopulations. Principle conclusions from these studies include: 1) If populations are connected by migration, uncontaminated sites cannot be reference sites. 2) The arrangement of patches is critical to overall impact of a toxicant. 3) If sufficient cleanup is not possible, it may be necessary to isolate the contaminated patch allowing formerly connected patches to regain more typical population dynamics.

Bio(s):
Louis Macovsky is a veterinarian (UGA, 76) with a wide range of practice experience (household pets, farm, and wildlife). With a lifelong interest in ecology, he returned to school and in 1999 received a MS (Environmental Science) at Huxley College of the Environment, WWU. It was during the latter that Louis was introduced to computer modeling. Since that time his primary interest has been exploring biological systems using computer modeling and simulation, particularly with the System Dynamics methodology. It is his intention to apply this approach of understanding and education to bring together stakeholders of current biological and ecological problems.

File attached or Link to Recording?
None



Date:
November 14, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Ed Ramsden

Title:
High Performace Discrete-System Simulation using Specialized Hardware

Abstract:
Common discrete-event (DE) simulation algorithms are both compute-intensive, and can be difficult to effectively partition for parallel execution on multiprocessors or other parallel computer architectures. By modeling the discrete-event system as a series of discrete-time state machines, it is possible to map certain types of DE models onto special-purpose electronic hardware - in effect a discrete analog computer. This computataional approach offers the potential for order-of-magnitude performance improvements over execution on general-purpose computer architectures.

Bio(s):
Ed Ramsden is currently an MS candidate in the Systems Science core program at PSU. Previously, he worked in the semiconductor and electronics manufacturing industries in a variety of technical and marketing positions, and holds a BSEE from Boston University

File attached or Link to Recording?
None



Date:
November 7, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Rod Walker

Title:
An Agent-Based Simulation of Stock Market Behavior

Abstract:
How much do investors really understand about how stock markets work? Even professional investment advisers over the last 20 years have generally recommended "buy and hold" strategies, implicitly admitting that they do not understand the markets very well themselves. Other investors believe they can do better using certain signals that tell them when to enter or exit the market. Traders buy and sell many times a day. To what extent do all of these different strategies actually affect the market itself? For example, if enough people follow a "buy and hold" strategy even Ponzi schemes will work -- until the scheme runs out of enough new investors. Stock markets are complex systems that are difficult to understand. Earnings seem to be important, but in the days of the dot-com boom, profits were almost seen as a bad thing. News events seem to be important, but stocks frequently rise on bad news and fall on good news. Investor sentiment seems to be important, yet markets are described as "climbing a wall of worry", with prices improving the most during times when most people think they won't. Given the complexity of the system, an agent-based simulation model has been created as a tool for exploring market behavior. This initial model will be presented for discussion. Rod Walker is a management consultant with a concentration in business dynamics. He holds BSEE and MBA degrees from the University of Texas at Austin. After 25 years as an engineer, manager, and executive at Texas Instruments and Compaq Computer, Mr. Walker left Compaq in 1998 to begin consulting. As a VP at Compaq, he participated in several important projects utilizing systems modeling experts with McKinsey & Company. The breakthrough nature of those projects prompted the inclusion of business simulation modeling as part of his emerging consulting practice

Bio(s):
Rod Walker is a current student in Systems Science, and is a management consultant with a concentration in business dynamics. He holds BSEE and MBA degrees from the University of Texas at Austin. After 25 years as an engineer, manager, and executive at Texas Instruments and Compaq Computer, Mr. Walker left Compaq in 1998 to begin consulting. As a VP at Compaq, he participated in several corporate-wide projects utilizing systems modeling experts with McKinsey & Company. The breakthrough nature of those projects prompted the inclusion of business simulation modeling as part of his emerging consulting practice.

File attached or Link to Recording?
None



Date:
October 31, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Jeff Fletcher

Title:
A Simple and General Explanation for the Evolution of Altruism

Abstract:
This talk presents a simple framework that highlights the most fundamental requirement for the evolution of altruism: assortment between individuals carrying the cooperative genotype and the helping behaviors of others with which these individuals interact. The framework decomposes fitness effects on individuals into those due to self and those due to the individual's ‘interaction environment’. For altruism to evolve, interaction environments experienced by altruists must be more generous than interaction environments experienced by non-altruists. This framework underlies, and is more general than, traditional explanations for the evolution of altruism (e.g. kin selection, multilevel selection, and reciprocal altruism). While kinship (genetic similarity) among those interacting is one way favorable interaction environments may be created, kinship is not a requirement for the evolution of altruism (as has been recently argued). In fact, even suicidal aid can theoretically evolve without help ever being exchanged among genetically similar individuals. This simple framework also helps clarify a common confusion made in the literature between alternative fitness accounting methods (which may equally apply to the same biological circumstances) and unique causal mechanisms for creating the assortment necessary for altruism to be favored by natural selection.

Bio(s):
Jeff Fletcher's research focuses on understanding the relationship among different theories on the evolution of altruism. He is also interested in developing ways to create more cooperative learning environments in the classroom and introducing undergraduate students to System Science ideas. He has a BS in Biology, an MS in Computer Science, and a Ph.D. in Systems Science. Jeff recently completed an NSF International Postdoctoral Fellowship at the University of British Columbia where he did research in the Department of Zoology and taught in the Integrated Science Program. He currently has a joint appointment in the University Studies Program and the Systems Science Graduate Program here at Portland State University.

File attached or Link to Recording?
None



Date:
October 24, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Tim Kochanski

Title:
A deeper understanding of iteration in simulations

Abstract:
I present an iterative model, programmed in Mathematica, which solves time paths for repeated Cournot games allowing us to see how output, price, profits, and market share in the 2-firm case change over time when one firm experiences per turn marginal cost reductions. By adjusting the marginal cost reduction rate for one firm and iterating, students can explore the various solutions and gain a better understanding of how the variables in the model diverge over time and the properties of that divergence. More generally, students gain experience designing models and programming in Mathematica and furthermore develop a deeper understanding of iteration in simulations.

Bio(s):
Formal Education and Project Background
Ph.D. Program Systems Science - Economics, Portland State University
M.S. Economics, U. of Oregon
B.A. Economics U. of Kansas
My primary academic interests include the history and practice of computational economics. The following presentation is based on a computational microeconomic project that I developed as an assignment in an undergraduate mathematics course. The theoretical model is common in undergraduate Industrial Organization texts today. While teaching at the University of Alaska Southeast I incorporated the project into the microeconomics course I was teaching and wrote up a paper. I modeled the paper after one that I had read in CHEER (Computers in Higher Education Economics Review). I presented the paper at a regional economics teaching conference, made a few changes, and submitted it to CHEER where it was recently published.

File attached or Link to Recording?
None



Date:
October 17, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dan Coates

Title:
What do we really understand about natural vision?
Controversy, consensus, and future directions in visual cognition, as informed by computational neuroscience.

Abstract:
The visual cortex is one of the most widely studied parts of the brain, and some believe it may be the key to cracking the so-called 'neural code.' Yet even after more than 50 years of intense scrutiny many mysteries remain. This talk offers an overview of the wide spectrum of fact and opinion in neural sensory processing.

Implementation of neural models, including recent computational studies, could provide insight into how we perceive and may help us understand the nature of cognition itself. Here particular attention will be paid to theories containing holistic notions, such as the Gestalt school of thought. It will be argued that only dynamic structured representations with intrinsic systematicity can accurately simulate neural function.

Bio(s):
Dan Coates is a Master's student in Computer Science at Portland State. He was recently an intern at Los Alamos Laboratory working with the PetaVision team building high-performance cortically-inspired models for visual object recognition. He is currently studying Gestalt perception in Melanie Mitchell's group, as well as working with Dan Hammerstrom on neurally-inspired hardware platforms.

File attached or Link to Recording?
None



Date:
October 10, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Rajesh Venkatachalapathy

Title:
RA and other Graphical Models

Abstract:
Graphical Models are one of the most used cook-book recipes in Machine Learning applications involving categorical data. After an introduction to Machine Learning and its connection with other fields (past,present,future),this talk will survey the connections between Reconstructibility Analysis and the more well known Graphical Models.

Bio(s):
Rajesh is currently a Systems Science student and is working with Martin Zwick on comparing RA with other Machine Learning Models.

File attached or Link to Recording?
None



Date:
June 6, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Scott Schecklman

Title:
Terahertz Rough Surface Scattering

Abstract:
"Terahertz research is one of the most intriguing and challenging fields to emerge in the 21st century. In less than a decade, this previously hidden section of the electromagnetic spectrum has caught the imagination of scientists around the world." ~ Gwo-Ching Wang, Physics Chair, RPI

Recent advances in laser technology have made it possible for scientists and engineers to access the far infrared portion of the electromagnetic spectrum. Within this so-called "THz gap" most non-polar packaging materials become transparent, while many other materials of interest have molecular resonances which can be used for spectroscopic detection and identification. This technology is proving to be useful in security screening and medical scanning for remote detection of drugs, explosives and even skin cancer.

However, at THz frequencies the penetration depth in hydrated materials, such as the human body, is quite limited. This restricts the usefulness of THz imaging to surface scans in a reflection arrangement. Furthermore, due to the very short wavelengths (hundreds of microns) many materials of interest appear to have random rough surfaces in this regime. Thus scattering of THz waves from rough surfaces threatens to corrupt THz spectroscopy and imaging measurements.

The Northwest Electromagnetic and Acoustics Research (NEAR) Lab at PSU is developing analytic and numeric algorithms to model electromagnetic wave scattering from random rough surfaces such as sandpaper. These models may one day be used to account for the scattering effect and recover the THz signature for material identification.

Bio(s):
Scott Shecklman received a Bachelor's degree in Electrical Engineering from Michigan Technological University in 1995. He worked for over 10 years as a radio frequency engineer designing and optimizing antenna systems and wireless telecommunications networks for Sprint, Qwest and US Cellular. Scott returned to school full time in 2007 and is currently pursuing a Masters degree in Electrical Engineering at Portland State University. He works as a research assistant for Professor Lisa Zurk in the Northwest Electromagnetic and Acoustics Research (NEAR) Lab and hopes to graduate in the fall of 2008.

File attached or Link to Recording?
None



Date:
May 30, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Tim Kochanski

Title:
An Agent Based Cournot Simulation with Innovation – Identifying the Determinants of Market Concentration

Abstract:
This paper uses an agent based Cournot simulation to study the effects of innovation on market concentration, as measured by a Herfindahl-Hirschman Index (HHI), under various market conditions. The model accommodates the following components: multiple firms with heterogeneous marginal costs, market entry and exit, barriers to entry, low or high cost industries, changing demand, varying levels of marginal cost reducing returns-to-innovation, varying costs associated with innovation, cost penalties for new entrants, increased returns to innovation from past experience innovating, and varying propensities to innovate within the market. The components mentioned above are commonly sited as determinants of market concentration.

To study the effects that each parameter has on market concentration a sensitivity analysis similar to that developed by previous research is employed (Brenner, Thomas 2001). Parameter ranges are specified based on economic theory and logical reasoning given the numeric values assigned to the intercept and slope of the demand function. All parameters are varied simultaneously except one which is varied systematically. At each setting of the systematically varied parameter a number of HHI means are collected in a sample. The systematically varied parameter is adjusted incrementally so that samples for a number of settings are generated.

The mean value of the HHI means is then collected for each sample. Via regression analysis, a line of best fit is used to estimate the effect of the systematically varied parameter on market concentration measured as the firm-adjusted HHI. A t-test is then performed to determine if the variation in HHI caused by each systematically varied parameter, given random variation in the other parameters, is statistically significant. The findings, which are consistent with economic theory, suggest that innovation in high cost industries leads to greater market concentration than does innovation in low cost industries, innovation under increasing demand leads to lower market concentration than does innovation under decreasing demand, and innovation under high barriers to entry lead to increased market concentration.

Bio(s):
Tim Kochanski was raised in a small college town in Kansas. After many trials, tribulations, and near-death experiences, he graduated from the University of Kansas with a B.A. in economics and made his way to Oregon arriving around the turn of the new millennium. He received his M.S. in economics from the University of Oregon in 2001 and started thinking about systems science as a possible doctoral program based on a buddy's recommendation. He worked many odd jobs through 2005 and audited many math classes at the U of O. In 2005 he took a part-time visiting professor position in Alaska and returned to Portland the following year to begin his Ph.D. in Systems Science - Economics.

File attached or Link to Recording?
None



Date:
May 23, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Katie McDonald, Dora Raymaker, Christina Nicolaidis

Title:
Participatory Action Research Strategies with Adults with Intellectual and Developmental Disabilities: Advancing Scientifically Sound, Socially Relevant, Ethical Research

Abstract:
To address negative attitudes towards individuals with intellectual and developmental disabilities (IDD), promote self-determination, and increase positive development, scientists have been called to embrace new research strategies. Participatory action research (PAR), which brings researchers and individuals with IDD into partnership, encourages science that recognizes the contribution of individuals with IDDs, focuses on issues of importance to individuals with IDDs, and better positions those individuals to improve their lives. Here, we discuss how principles of PAR foster scientifically sound, socially relevant, ethical research aimed at promoting positive social change. We will provide examples from an autistic-academic researcher partnership, the Academic Autistic Spectrum Partnership in Research and Education (AASPIRE). We will also relate PAR and AASPIRE with concepts from Senge's _The Fifth Discipline_, the value of multiple perspectives, and ecological models of research.

Bio(s):
None

File attached or Link to Recording?
None



Date:
May 16, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dr. Martin Zwick

Title:
Systems Metaphysics: A Bridge from Science to Religion

Abstract:
"Systems theory" is familiar to many as the scientific enterprise that includes the study of chaos, networks, and complex adaptive systems. It is less widely appreciated that the systems research program offers a world view that transcends the individual scientific disciplines. We do not live, as some argue, in a post-metaphysical age, but rather at a time when a new metaphysics is being constructed. This metaphysics is scientific and derives from graph theory, information theory, non-linear dynamics, decision theory, game theory, generalized evolution, and other transdisciplinary theories. These 'systems' theories focus on form and process, independent of materiality; they are thus relevant to both the natural and social sciences and even to the humanities and the arts. Concerned more with the complex than the very small or very large, they constitute a metaphysics that is centered in biology, and thus near rather than far from the human scale.

Systems metaphysics forges a unity of science based on what is general instead of what is fundamental; it is thus genuinely about everything. It counters the nihilism of narrow interpretations of science by affirming the link between fact and value and the reality of purpose and freedom in the natural world. It offers scientific knowledge that is individually useful as a source of insight, not merely societally useful as a source of technology. With the new world view that it brings, systems metaphysics contributes to the recovery of cultural coherence. It builds a philosophical bridge between science and religion that is informed by our understanding of living systems. It suggests a secular theodicy in which imperfection is lawful yet perfecting is always possible, and uses this perspective to analyze religions as systems. It provides scientific insights into traditional religious concepts, including those ideas that guide spiritual practice.

Full paper at: http://www.metanexus.net/magazine/tabid/68/id/10051/Default.aspx
Related papers at: http://www.pdx.edu/sysc/research_systemsphilosophy.html

Bio(s):
Martin Zwick was awarded his Ph.D. in Biophysics at MIT in 1968, and joined the Biophysics Department faculty of the University of Chicago in 1969. Initially working in crystallography and macromolecular structure, his interests shifted to systems theory and methodology, the field now known as the study of chaos, complexity, and complex adaptive systems. Since 1976 he has been teaching and doing research in the Systems Science PhD Program at Portland State University; during the years 1984-1989 he was director of the program.

His main research areas are information theoretic modeling, machine learning, theoretical biology, game theory, and systems theory and philosophy. Scientifically, his focus is on applying systems theory and methodology to the natural and social sciences, most recently to biomedical data analysis, the evolution of cooperation, and sustainability. Philosophically, his focus is on how systems ideas relate to classical and contemporary philosophy, how they offer a bridge between science and religion, and how they can help us understand and address societal problems.

File attached or Link to Recording?
None



Date:
May 2, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dr. George G. Lendaris

Title:
Some Ideas on How to Achieve Human-Like Experience-Based Control via Computational Agents – and Lots of Unanswered Questions

Abstract:
Distinguishing features of human-like control vis-à-vis current technological control include the abilities to make use of experience while selecting control policies for distinct situations, and to do so faster and faster as more experience is gained. The latter is in stark contrast to current technological implementations that slow down as more knowledge is stored. How do we humans do it? The notions of context and context discernment are posited to be important stepping stones to understanding and implementing these human abilities.

Whereas methods known as Adaptive Control and Learning Control focus on modifying the design of a controller as changes in context occur, experience-based control (EB) entails selecting a controller design from a collection of designs previously developed for similar contexts. Developing the EB approach entails a shift of the technologist's focus "up a level", away from designing individual (optimal) controllers to that of developing on-line algorithms that efficiently and effectively select designs from a repository of existing controller solutions, previously populated via "experience".

A notion to be described is that of Higher Level Learning Algorithm. This is a new application of Reinforcement Learning -- Approximate Dynamic Programming (ADP) variety -- with its focus shifted to the posited higher level. Some promising examples will be described.

Bio(s):
George G. Lendaris is Professor of Systems Science and Electrical & Computer Engineering at Portland State University.

File attached or Link to Recording?
None



Date:
April 25, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dr. Radu Popa

Title:
Fundamental asymmetries and the origin of homochiral order

Abstract:
Attached

Bio(s):
Dr. Popa is an associate professor of microbiology at Portland State University. His research projects include Microbial Ecology, Geomicrobiology, and Astrobiology. He previously worked as a research assistant professor at USC, and a post doctoral fellow at the Jet Propulstion Laboratory at Caltech. He is the author of the book, "Between Necessity and Probability: Searching for the Definition and Origin of Life."

File attached or Link to Recording?
.pdf


 


Date:
April 18, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Edward Ramsden

Title:
Simulation with Distributed Variables

Abstract:
Physical systems with spatially distributed variables are widespread in scientific and engineering models. Examples of such systems include electro-magnetic fields and mechanical vibrations. Modeling and simulating these systems requires techniques that are significantly different from than those used for either traditional systems dynamics or discrete event systems. This seminar will describe some of the techniques for formulating models of these types of systems and simulating them.

Bio(s):
Ed is currently working toward an MS in Systems Science. Before that, he spent twenty years working in the field of electronic engineering.

File attached or Link to Recording?
None



Date:
April 11, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dr. Steve Bleiler

Title:
An Introduction to the Quantum Theory of Games

Abstract:
Computers and networks that exploit the bizarre properties of quantum mechanics will have capabilities far exceeding those of the conventional computing environment. The encryption of data, the searching databases, and even the play of simple games such as on-line poker will undergo profound changes when implemented in the quantum environment.

This is because players who communicate their strategic choices via quantum channels can, in effect, put their strategic choices in superposition, and thus have access to a vastly larger selection of strategic choices than that available to players communicating via classical channels.

For some simple games, it is enough that one player have access to these quantum strategies when the other does not to ensure the first player's certain victory. Yet for most two-player games, mere access to quantum strategies is merely an expensive way to implement what game theorists call mixed strategies. Strategic choices in a mixed strategy are determined randomly with specific probabilities by the individual players. Accessing the larger collection of quantum strategies in this instance requires the utilization of yet another strange phenomenon of the quantum world, that of entanglement. In the entangled version of a given game new "solutions" to the game present themselves that perform better than the "solutions" available to players of the classical version.

The talk will begin with a brief review of the relevant axioms of quantum mechanics. After a short discussion of what is meant by a "quantization" of a game, we will consider some simple examples, D. Meyers original "penny flip" game and a specific entangled protocol, originally developed by Eisert, Wilkens and Lowenstein, for quantizing simple games via mediated quantum communication. If time allows, by appealing to a quaternionic representation of the EWL protocol developed by S. Landsburg, we'll illustrate the final sentence of the previous paragraph for a simplified form of poker, where "quantum" bluffing is always more profitable than bluffing "classically", and even is profitable when classical bluffing is not!

Bio(s):
As for me: I received my Ph.D. in Mathematics from the University of Oregon in 1981, have held positions at the Universities of Texas, Utah, British Columbia, and Melbourne (Australia) in addition to my Professorship at PSU where I have been since September of 1988. I am the author of nearly 30 scientific articles and books (some of which have been translated into Russian), have a knot named after me, and am an accomplished juggler, poker player, x-country skier, and mountaineer. In 2003 I was the Mathematical Association of America's Distinguished Teacher for the Pacific Northwest and I hold a John Elliot Allan Award for Distinguished Teaching here at PSU. Classically trained as a topologist, I have published papers in topology, geometry, combinatorics, group theory, chaos, fractals , and solotons, and now game theory. In my spare time, I competed in the Championship event of the WSOP in 2003 and 2007 (finishing 297th out of 839 in 2003 and 2145th out of 8757 in 2007) playing my way in through the satellite system in 2003 and on-line in 2007.

My lectures on the game theory of poker at places such as the University of Nevada-Las Vegas, the University of Illinois-Chicago, the University of Washington, and Oregon State University, have played to sold out crowds. I teach a course here at PSU on the mathematics of poker (next offered this coming summer) and the notes for which are scheduled to be published in book form in the Springer-Verlag Undergraduate Texts in Mathematics Series.

File attached or Link to Recording?
None



Date:
April 4, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Joe Fusion

Title:
Python for the Systems Scientist

Abstract:
In this seminar, I will present an overview of the Python programming language. Python is a high-level, general-purpose language, available on many platforms. It features a large standard library of useful extensions, and the ability to connect to many other languages and protocols. These characteristics, and many others, make Python a highly useful tool for a wide variety of research tasks. My goal will be to familiarize you with these characteristics, rather than to teach programming. I'll show some examples, including how Python is used with Dr. Zwick's Occam software, and perhaps some other data processing and presentation techniques.
Bio(s):
Joe Fusion is a Ph.D. student in the core Systems Science program. His research interests include artificial life, theoretical biology, systems modeling, and philosophy.

File attached or Link to Recording?
None



Date:
March 14, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
John Anasis
Bonneville Power Administration

Title:
Power System Congestion Management:
The Balance Between Reliability, Equity, and Economics

Abstract:
Load growth, the addition of new generation, and the lack of new transmission infrastructure over the past 10-20 years have placed an ever increasing strain on the electric power grid. Grid operators are finding that flows on the system are approaching or exceeding reliability limits more frequently. This loading of the power system up to or past its reliability limits is known as “congestion”. Grid operators have several competing criteria they must balance in order to manage congestion. They must take actions that preserve the reliability of the power system; however, they must do so in a manner that does not unfairly discriminate between participants in the power market. Their actions must also not result in large financial costs to all users of the grid.

Bio(s):
John Anasis received his B.S. in Electrical Engineering with a minor in Physics from the University of Portland in 1985 and a Masters in Public Administration from Portland State University in 1989. He is currently a student in the Systems Science Ph.D. program at Portland State. John joined the Bonneville Power Administration in 1985 as an electrical engineer in the Remedial Action Scheme design section of BPA's Control Engineering Branch. He has since held several positions with BPA's System Operations and Transmission Marketing groups where he has performed a wide range of duties, including power system analysis, development of operating instructions, determination of available transmission capacity for sale, tariff and business practice development, and the review of power industry restructuring issues. He is currently with BPA's Technical Operations Branch where his primary duties are the determination of safe operating limits for the BPA transmission grid.

File attached or Link to Recording?
None



Date:
March 7, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Sharon Glaeser & Andrew Toland

Title:
Neural Networks and Call Recognition in Elephants

Abstract:
Relatively little is known about the vocal repertoire of Asian elephants. A categorization of basic call types and modifications of these call types by quantitative acoustic parameters is needed to examine acoustic variability within and among call types, to examine individuality, to determine meaning of calls via playback, and to develop rigorous call recognition algorithms for acoustic monitoring and census of wild populations. By studying communication of known individuals in the more controlled setting offered by captivity, more rigorous analyses can be done with regards to individuality, social context, variability, reproductive state, and perceived emotional state. This project aims to 1) categorize sounds by acoustic parameters, 2) define an acoustic repertoire of captive Asian elephants, 3) examine functional relevance of acoustic variability in a captive environment, 4) investigate individuality, 5) examine the function of low-frequency communication in captivity to determine potential impact of low-frequency anthropogenic noise, and 6) develop call recognition and potentially individual recognition algorithms for the Asian elephant acoustic repertoire. Through collaboration with Andrew Toland in Systems Science, neural networks are being used for call recognition by classifying elephant vocalizations from within a time series containing other acoustic events and background noises. Starting from a spectrographic representation of the vocalizations, a reduced representation is obtained using an autoassociative network. Following this, a recurrent neural network is trained to recognize the dynamic patterns associated with specific vocalization types. Early results indicate that the suggested architecture succeeds in classifying approximately 90% of the samples in the test set.

Bio(s):
Sharon Glaeser is a masters student in the Department of Biology at Portland State University. For her masters research she is studying acoustic communication in the Oregon Zoo's Asian elephants. In short, she aims to define an acoustic repertoire for Asian elephants, to investigate variability and individuality, to develop a call recognition algorithm through collaboration with Andrew Toland in Systems Science, and to provide a basis for comparisons between captive and wild Asian elephants and between Asian and African elephants. Her major advisor at PSU is Dr. Randy Zelick.

Andrew Toland is a graduate student in Systems Science, and is interested in neural networks.

File attached or Link to Recording?
None



Date:
February 29, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Dan Hammerstrom
ECE Department, PSU

Title:
The Cortical Algorithm As A Bayesian Network - Some Speculation

Abstract:
The semiconductor industry has been following Moore's law for over 40 years, enabling a revolution in computing that has had huge societal and industrial impact. The industry is now designing and manufacturing transistors that are in the 45 nanometer range, in chips that have over 1 billion transistors.

However, in spite of all this processing speed and cheap computing power, we still have not solved the hardest problems in computing, making computers more intelligent.

Motivated by the fact that biological systems have successfully dealt with similar issues, a number of researchers are beginning to look at biological models, primarily those from computational and systems neuroscience, and cognitive science, for inspiration for new chip architectures that are a better match to the molecular scale electronics.

A number of neuroscientists are focused, in particular, on cerebral cortex. Nature has, so it appears, produced a general purpose computational device in cortex that is a fundamental component of higher level intelligence. Although we are a long ways from understanding the details of how cortex works, some of the basic computations are beginning to take shape.

Preliminary speculation of the cortical algorithm assumes a modular structure, where each module is implemented by a simple associative network that does a kind of Bayesian inference. The modules are then organized in a 2D layout with sparse inter-module interconnect.

In this presentation I introduce the Bayesian Memory, which is loosely based on ideas and principles from the current, limited, understanding of cerebral cortex. We believe that the Bayesian Memory is a rough first step in creating a generalized building block for Intelligent systems and which has a clean mapping to hybrid CMOS / nano-electronic implementations.

Bio(s):
Dan Hammerstrom is Professor and Associate Dean for Research in the Electrical and Computer Engineering Department at Portland State University.

File attached or Link to Recording?
None



Date:
February 15, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Marek Perkowski

Title:
Engineering Introduction to Quantum Computing

Abstract:
The overview talk will present research on Quantum Computing performed by the group of Dr. Perkowski in Electrical and Computer Engineering Department. It will include the presentation of basic quantum computing concepts, gates and circuits. The second part will present our research in synthesis of binary and multiple-valued quantum circuits, testing quantum circuits, quantum algorithms and specifically the Grover-based solving of combinatorial problems, especially from electronics CAD. Grover algorithm gives quadratic speedup on every NP problem provided that we can design a quantum oracle for it.

Bio(s):
Dr. Perkowski is Professor of Electrical and Computer Engineering at Portland State University. He is a member of the Portland Quantum Logic Group and the Portland Logic and Optimization Group. He is also the Director of the Intelligent Robotics Laboratory at PSU. His interests include many aspects of machine learning, programming, and teaching. He teaches courses in Quantum Computing, Intelligent Robotics, Advanced Logic Synthesis, Robot Vision and Perception, among others.

File attached or Link to Recording?
None



Date:
February 8, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Andrew Toland

Title:
Discussion of Mathematica

Abstract:
The seminar for Friday, Feb. 8 will be canceled due to an unfortunate scheduling misshap.
However, for those who are interested, Andrew Toland will continue his discussion of Mathematica by presenting several examples from class projects. The presentation will be ad-hoc, so there is not an abstract. The examples will come from numerical math and engineering courses mainly.

Bio(s):
None

File attached or Link to Recording?
None



Date:
February 1, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Jeff Fletcher

Title:
Evolution of Altruism Theory: Different Accounting Methods or Different Causal Explanations?

Abstract:
For several decades the mechanisms by which altruistic and cooperative behaviors evolve have been vigorously debated. The main theories are kin selection (or inclusive fitness) theory, reciprocal altruism theory (including variations based on reputation, sanctions, and spatial structure), and multilevel (or group selection) theory. This debate has recently intensified in the literature with publications emphasizing the role of group selection in the evolution of eusociality and several articles in the last year that claim kin selection (genetic similarity between altruists and recipients) is the only mechanism that can account for biological altruism. My work has focused on unifying different theories of how altruism evolves and my talk will consider this recent controversy and its history from this perspective, including how these theories vary in their definitions of altruism and differences in the way they keep track of fitness consequences.

Bio(s):
Jeff Fletcher has a BS in Biology, an MS in Computer Science, and designed medical records software for 7 years. He completed his Ph.D. in Systems Science from Portland State University in 2004. He recently completed an NSF International Postdoctoral Fellowship at the University of British Columbia, Department of Zoology, where he also taught in the Integrated Science Program. His research has focused on understanding the relationship among different theories on the evolution of altruism. Currently Jeff teaches for the University Studies and System Science Ph.D. Programs here at Portland State.

File attached or Link to Recording?
None



Date:
January 25, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Robin Fenske

Title:
An Prisoner's Dilemma Solution in Agent Based Simulation

Abstract:
This model explores the problem of the Prisoner's Dilemma using Agent Based Simulation without the traditional solution of iterations with the same partner (Axelrod 1984).

Agent-based models differ from most computer models in that the computation is decentralized, not centralized. Each individual agent can have variables associated with it, instead of having variables representing the aggregate properties of the system. These variables can change as the agents move and interact with their environment. Agents can be identical or they can be of different 'breeds.' One can specify behaviors and decision-making rules for a each breed of agent and control each breed separately. The aggregate behavior "emerges" from the interaction of the agents and the environment. (From the class description). ABS is offered as SySc 525/625.

In an attempt to make the standard Prisoner's Dilemma model closer to real world community interactions, noniterative interactions were used, and methods of increasing cooperation in this setting were developed. Non-traditional solutions to the Prisoner's Dilemma have presented before, such as voluntary re-partnering (Joyce, Kennison, Densmore, Guerin, Barr, Charles, and Thompson, 2006) and (Boone & Macy, 1999), interpersonal commitment (Back & Flache, 2006), and reliance on personal experience (Fort, 2003). The model presented here is unique from the articles mentioned above. * *The agents in this model do not have an assigned multi-round pattern of behavior, nor do they have any memory of their own experience or knowledge of other agent's experiences.

Bio(s):
Robin Fenske is a full time first year PhD student in the Systems Science Graduate Program. She is interested in applying Systems Science concepts to Sustainability, Local Economies, and Human Decision Making. She holds a Bachelor of Science from The Evergreen State College. She is also working with Dr. Wayne Wakeland on a "Food Delivery Carbon Foodprint" inter-departmental research grant.

File attached or Link to Recording?
None



Date:
January 18, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Andrew Toland

Title:
An Introduction to Mathematica

Abstract:
This week we'll take an introductory look at Mathematica, another commercially available math package. For students, I think it's well worth the price and the time to learn it. If I had to choose one piece of software to take with me to a desert island, this might be it (and I might need some time alone on a desert island to fully learn it). I hope to show the basics of interacting with the "front end" and enough syntax to feel comfortable getting started. We'll look at version 6, the latest version. It looks like it has some really nice new features for direct interactivity and visualization. Also, there has been a major revision in the help browser that should make the process of learning the software more accessible.

Bio(s):
Andrew Toland is a graduate student in the Systems Science department. He's had opportunities to use Mathematica in school and on the job (in the rare instances when he's actually had a job).

File attached or Link to Recording?
None



Date:
January 11, 2008

Location:
Harder House, Room 104

Time:
12-1 pm

Presenter(s):
Lars Holmstrom

Title:
Demystifying the Matlab Computing Environment

Abstract:
Matlab is a software computing environment that is becoming more and more popular across a number of disciplines, including finance, biology, statistics, and, yes, Systems Science. While it may appear at first to be a high powered graphic calculator, its extensibility and toolbox expansions allow it to be much more. Unfortunately, this "everything" tool seems prohibitively complicated to many who may find it very valuable. In this seminar, I will be giving a tutorial on how to get your feet wet in the world of Matlab. All are welcome, but my target audience is people who have never heard of Matlab, have been curious about it, or are just getting started on this exciting path. I won't be focusing on individual toolboxes or techniques, but will rather make use of the time to demystify the startup phase and to provide people with the tools and direction for learning more on their own.

Bio(s):
Lars Holmstrom is a PhD student in the Systems Science Program at Portland State University. He often ponders what his life as a mathematical modeler would be like without Matlab and doesn't like where his mind wanders.

File attached or Link to Recording?
None