|
Date: Tuesday July 14, 2009
Speaker: Professor Andrew Lo (MIT Sloan
School)
Location: The MITRE Corporation (Bedford, MA &
McLean, VA)
Title: "Kill All the Quants"?: Models vs.
Mania in the Current Financial Crisis
Abstract:
As the shockwaves of the financial crisis of 2008
propagate throughout the global economy, the "blame game" has
begun in earnest, with some fingers pointing to the complexity of certain
financial securities, and the mathematical models used to manage them. In
this talk, I will review the evidence for and against this view, and argue
that a broader perspective will show a much different picture. Blaming
quantitative analysis for the financial crisis is akin to blaming F = MA
for a fallen mountain climber's death. A more productive line of inquiry is
to look deeper into the underlying causes of financial crisis, which
ultimately leads to the conclusion that bubbles, crashes, and market
dislocation are unavoidable consequences of hardwired human behavior
coupled with free enterprise and modern capitalism. However, even though
crises cannot be legislated away, there are many ways to reduce their
disruptive effects, and I will conclude with a set of proposals for
regulatory reform.
Date: Wednesday, April 22, 2009
Speaker: Professor Venkatesh Saligrama, Boston University
Location: Verizon Laboratories (Waltham, MA)
Title: Video Analytics over Camera Networks
Abstract:
Network of video cameras, developed in the last decade
or so, permit today pervasive, wide-area visual surveillance. The main
difficulty is that video data by its nature produces a high degree of
clutter and it is difficult to identify the truly relevant information from
clutter particularly in urban environments. Conventionally, this problem
has been handled by an object-based approach wherein an object may first be
tagged, identified, classified, and tracked before behavior modeling and
abnormal detection.
This paradigm is neither scalable to complex urban
environments or to networks of video cameras having limited communication
capability.
We explore a location-based approach for behavior
modeling and abnormality detection. We proceed directly with event
characterization and behavior modeling at the pixel(s) level based on
motion labels obtained from background subtraction. Our method requires
little processing power and memory, is robust to motion segmentation
errors, and general enough to monitor humans, cars or any other moving
objects in uncluttered as well as highly-cluttered scenes.
Date: Wednesday, February 25, 2009
Speaker: Prof. Dominique Haughton
Location:
Emptoris, Inc. (Burlington, MA)
Title: Multilevel models and living standards in
Vietnam
Abstract:
This talk describes an investigation of living
standards in Vietnam
from the point of view of a study of real household per capita expenditure
by means of two multilevel models. Policy implications of this work for
poverty targeting will be discussed as we go along.
*First we will give an introduction to the Vietnam
Household Living Standards Surveys (VHLSSs).
*We will then demonstrate how the use the software
MLWin to construct a multilevel model with the logarithm of real per capita
expenditure as a dependent variable, and a number of household
characteristics such as age and education of the head of household, etc.
and an urban/rural location indicator as independent variables. The model
includes four nested levels, the household (level 1) inside a commune
(level 2) inside a district (level 3) inside a province (level 4).
*We will discuss how the model yields random effects
in the intercept that provide an estimate of differences in living
standards due to geographical location, when household characteristics are
controlled for. We will also suggest that random effects identified in the
independent variable urban/rural can be seen as location-specific random
contributions to the urban/rural gap above and beyond the effects of known
location characteristics, such as level of education of the population,
etc.
*We then demonstrate how the multilevel model can be
used to obtain small area estimates of the mean (logarithm of) household
per capita expenditure at the commune level.
*Further current perspectives will then be discussed,
regarding an extension of our model to accommodate data from three Vietnam
Household Living Standards Surveys, of 2002, 2004 and 2006.
Date: Wednesday, October 29, 2008
Speaker: Prof. Karl Rexer
Location:
Emptoris, Inc. (Burlington, MA)
Title: Data Miners' Views on Data Mining: Survey
Findings & One
Consultant's Perspective
Abstract:
Results of the 2nd Annual Rexer Analytics Data Miner
Survey will be presented. The survey of 348 data miners examined the
algorithms and tools currently being employed, priorities considered in
selecting these tools, the types of data analyzed, challenges encountered,
and solutions provided. Trends between the 2007 and 2008 findings will be
highlighted.
Additionally, Karl will discuss his perspectives on
data mining software and algorithms, and will provide tips for achieving
success in data mining projects.
Date: Wednesday, October 1st, 2008
Speaker: Prof. DIMITRIS BERTSIMAS
Location: Verizon Network and Technology (Waltham, MA)
Title: OR in health care: opportunities, perspectives
and progress
Abstract:
In recent years the availability of massive amounts
of electronically available data involving millions of people and the
development of new data mining algorithms present an exciting new
opportunity for Operations Research to have a significant impact in health
care.
We discuss our research efforts in assessing the risk
of patients, their quality of care and also propose a new data based
assessment for cancer research. We further discuss further research
directions.
Date: Tuesday, June 10th, 2008
Speaker: Prof. Alfred Blumstein
Location: Emptoris, Inc. (Burlington, MA)
Title: Bringing OR Perspectives to the Criminal
Justice System
Abstract:
OR practitioners are likely to find the CJS one of
the most primitive of social systems. Over the past forty years, I and a
number of colleagues have been involved in bringing OR techniques of
quantitative modeling, system perspective, and planning to that system The
issues addressed have included modeling of criminal careers as a stochastic
process, bringing those analyses to the assessment of incapacitation
effects of incarceration, review of trends in incarceration and factors
contributing to those trends. We specifically focus on the interaction of
incarceration and drug markets and identify some of the unintended
counterproductive effects of that response to the drug problem.
Note that the slides and an audio recording of the
talk are available for Boston Informs chapter members. Please contact the
webmaster for details on how to retrieve the files.
Date: Wednesday, April 30th, 2008
Speaker: Prof. Arnold Barnett
Location:
Emptoris, Inc. (Burlington, MA)
Title: The Electoral College: Its Care and Cure
Abstract:
As the 2000 election so vividly showed, it is
Electoral College standings rather than national popular votes that
determine who becomes President. But how do we understand what is happening
under the Electoral College rules? When a poll reports that (say) Clinton would have 45%
of the votes in a Clinton-McCain contest and McCain 43%, which one is ahead
in the Electoral College? And what is the probability that this candidate
is genuinely ahead, taking into account the statistical sampling error that
attends each statewide result?
Some people think that the Electoral College should
be replaced by a national popular vote. However, that will not happen. We
will discuss a modification of current procedures that would come very
close to a national popular vote, yet actually increase the role of the
smaller states that benefit from present arrangements. Conceivably, the
system could be in place by 2012. Not that one should bet on it.
Date: Wednesday, November 28, 2007
Speaker: Prof. David Jensen
Location: Verizon Laboratories (Waltham, MA)
Title: Learning and exploiting statistical
dependencies in networks
Abstract:
Networks are a ubiquitous representation for natural,
technological, and social systems. We live embedded in social and
professional networks, we communicate through telecommunications and computer
networks, and we represent information in documents connected by hyperlinks
and bibliographic citations. Only recently, however, have researchers
developed techniques to analyze and model data probabilistic dependencies
in these networks. These techniques build on work in artificial
intelligence, statistics, databases, graph theory, and social network
analysis, and they are profoundly expanding the phenomena that we can
understand and predict. However, new frontiers await.
In this talk, I will survey some recent work in
learning probabilistic models of relational data, and discuss several
applications of these techniques, including fraud detection in the U.S.
securities industry. I will argue that current techniques are capable of
learning only a subset of the knowledge needed by practitioners in these
domains, and that further work in analyzing networks offers a unique
ability to produce the full range of knowledge needed in a wide range of
applications, including a unification of work in machine learning, causal
inference, and agent-based simulation.
Date: Thursday, October 11, 2007
Speaker: Prof. Karla Hoffman
Location: Riverfront
Conference Center
(Cambridge, MA)
Title: The Dance Of The Thirty-Ton Trucks:
Dispatching And Scheduling In A Dynamic Environment
Abstract:
We report on the application of operations research
to a very complex scheduling and dispatching problem. Scheduling and
Dispatching is never easy, but the scheduling of concrete deliveries is
particularly difficult for several reasons: 1) concrete is an extremely
perishable product – it can solidify in the truck if offloading is
delayed by a few hours; 2) customer orders are extremely unpredictable and
volatile – orders are often canceled or drastically changed at the last
minute; 3) the concrete company overbooks by as much as 20% to compensate
for customer unpredictability; 4) many orders require synchronized
deliveries by multiple trucks; 5) when a truck arrives at a customer site,
the customer may not be ready for the delivery or a storm may negate the
ability to use the concrete; and 6) most of the travel takes place in
highly congested urban areas making travel times highly variable.
In order to assist the dispatchers, schedulers, and
order-takers at this company, we designed and implemented a decision
support tool consisting of both planning and execution tools. The modules
determine whether new orders should be accepted, when drivers should arrive
for work, the real-time assignment of drivers to delivery loads, the
dispatching of these drivers to customers and back to plants, and the
scheduling of the truck loadings at the plants.
For the real-time dispatching and order-taking
decisions, optimization models are solved to within 1% of optimality every
5 minutes throughout the day. This nearly continuous re-optimization of the
entire system allows quick reactions to changes. The modeling foundation is
a time-space network with integer side constraints. We describe each of the
models and explain how we handle imperfect data. We also detail how we
overcome a variety of implementation issues.
The success of this project can be measured, most
importantly, by the fact that the tool is being ported by the parent
company, Florida Rock, to each of its other ready-mix concrete companies.
Secondly, the corporation is sufficiently convinced of its importance that
they have begun promoting this methodology as a "best practice"
at "World of Concrete" and "ConAgg" industry
conventions.
Date: Tuesday, Sept, 18, 2007
Speaker: Prof. Cynthia Barnhart
Location:
Emptoris, Inc. (Burlington, MA)
Title: Airline Optimization: If we are so good at it,
why are things so bad?
Abstract:
Airlines have a history of OR work and have spent a
lot of money optimizing their schedules. So why do we keep hearing about all
the problems at the airports? There is still some work to do. This talk
covers some problem areas, and some proposed solutions.
Date: Wednesday, June 13, 2007
Speaker: Prof. Yossi Sheffi
Location: Verizon Laboratories (Waltham, MA)
Title: The Resilient Enterprise
Abstract:
What happens to a company when the unimaginable
occurs? When an earthquake hits its primary contract manufacturer? When
labor strikes shut down an entire port? When terrorists cripple a
transportation system?
In this talk, Yossi Sheffi, Professor of Engineering
at MIT and Director of the MIT
Center for
Transportation and Logistics, describes how a company’s survival and
prosperity depend more on what it does before such a disruption occurs than
on the actions it takes as the event unfolds. The talk is based on his
award-winning best seller "The Resilient Enterprise: Overcoming
Vulnerability for Competitive Advantage" (see
http://resilient-enterprisemit.edu/) which was reviewed by dozens of
publications, including the WSJ, NYT, and The Economist. The FT called in
one of the best business books of 2005.
The talk explores high-impact/ low-probability
disruptions, focusing not only on security but on corporate resilience¬the
ability to bounce back from such disruptions¬and how resilience investments
can be turned into competitive advantage. It demonstrates the concepts with
dozens case studies and examples.
Date: Tuesday May 8, 2007
Speaker: Thomas Leighton
Location:
Akamai Cambridge (Cambridge, MA)
Title: The Akamai Story: From Theory to Practice
Abstract:
Akamai Technologies is the leading global service
provider for accelerating content and business processes online. Prof.
Leighton will describe Akamai's origins and business. The talk will be
followed by a tour of Akamai's Network Operation & Control Center
(NOCC).
Date: Tuesday March 20, 2007
Speaker: Dr. John S. Hammond, co-author of Smart
Choices, A Practical Guide to Making Better Decisions
Location: Verizon Laboratories
Title: The Hidden Traps in OR/MS
Abstract:
As OR/MS professionals our models often require
subjective numerical inputs and subjective adjustment of outputs in the
course of the user making decisions. How good are these judgments? It turns
out that they are subject to ingrained psychological biases that can
sabotage the decisions, sometimes destroying much of the benefit of an
otherwise fine model.
Most decision makers and indeed many OR/MS
professionals are blissfully unaware of these traps, in spite of the fact
that Daniel Kahneman won the 2002 Nobel Prize in Economics for his
ground-breaking work on them, work reflected in this talk.
Our speaker, John Hammond, will introduce us to some
of the most important traps, illustrate where they occur, and suggest ways
to reduce their impact. The session, based in part on his best-of-HBR
Harvard Business Review article, “The Hidden Traps in Decision
Making,” will be highly interactive.
Date: Thursday, October 26th, 2006
Speaker: Dr. Daniel Bienstock, Professor of
Operations Research at Columbia
University
Location: Emptoris, Inc.
Title: Catastrophic blackouts: problems, models,
solution methods
Abstract:
The August, 2003 blackout that affected the Northeast
of the US and Canada, and similar events in other countries, have brought
attention on the fact that national power transmission networks have
become, over the past few decades, large, complex, and exposed to
catastrophic failures. From an OR perspective, the problem is complicated
for two reasons: first, power networks follow complex laws of physics that
are not easy to model, and second, blackouts are initiated by combinations
of exogenous events that are extremely unlikely even as they prove
damaging. In this talk we will
describe ongoing work to tackle this problem.
Date: Tuesday, Sept. 26, 2006
Speaker: Dr. Tom Magnanti, Dean of the School of
Engineering at MIT and former President of INFORMS
Location: Verizon Laboratories
Title: When OR is AND: Perspectives on Theory and
Practice in Operations Research
Abstract:
When is it possible (desirable) to bring together
theory AND practice, to have both rigor AND relevance? Unlike some fields
that place emphasis on one extreme or the other, Operations Research has
historically been motivated in large part by both. This talk will reflect
on the interplay between theory and practice – OR is AND when OR is
Operations Research. It will focus on applications of operations research
including several in engineering that might not be familiar to an
operations research audience.
Date: Thursday, Aug 3, 2006
Location: MIT
Lincoln
Laboratories
Speaker:
Colonel Darrall Henderson,
Title:
Operations Research and West Point
Supporting a Nation at War
Abstract:
This talk will share insights about the speaker's
tour in the Multinational Force - Iraq Headquarters. Specifically, he will
discuss analysis supporting the four lines of operation in the Campaign
Plan for OIF (Security, Governance, Economic, and Information). The pace of
planning and combat operations in Iraq requires one to quickly
collect, interpret and assimilate data and information to aid combatant
commanders. In order to accommodate these unique requirements, an eclectic
mix of operators, strategists, planners, analysts, and civilian subject
matter experts reminiscent of the early days of Operations Research is
required. Additionally, the speaker will discuss ongoing efforts at the United States Military
Academy at West Point, NY
to provide analysis to support a Nation at War.
Date:Tuesday, June 6, 2006
Location: Verizon
Laboratories
Speaker:
Dr. Mitch Burman,
Title:
Engineering Increases HP Productivity by $780M
Abstract:
As HP installed a system for manufacturing ink-jet
printers, it discovered that it was only going to perform at 20% of
anticipated output levels. After several unsuccessful attempts to use standard
software tools, we determined the need to use a first principles industrial
engineering approach. As a result, we developed an analytical model of the
production line dynamics that allowed HP to achieve the desired results.
The experience demonstrated that there was no substitute for a rigorous
engineering approach. This talk will focus on the HP case study, but also
touch upon the pitfalls of relying too much on software to solve industrial
operations challenges.
Date: Wednesday, May 3, 2006
Location: Emptoris, Inc.
Speaker:
Dr. Nitin Patel,
Title:
Introduction to Data Mining
Abstract:
This talk will introduce the rapidly evolving,
multidisciplinary field of data mining. Data mining is rooted in the fields
of statistics and artificial intelligence and has been gaining momentum in
managerial applications in areas such as customer relationship management,
finance, e-business, and operations. Data mining seeks to exploit the large
volumes of transactional data that are routinely stored in databases to
record operational steps such as moving, storing, testing, and selling
products or delivering services like financial credit, maintenance and call
center support.
As is often the case with emerging fields having
substantial commercial potential, there is considerable hype concerning
data mining. My aim is to enable a balanced view of data mining for
management scientists interested in harnessing this technology to improve
business decisions. I will provide an overview of the types of problems
addressed by data mining and the key ideas that underlie its approach. I
will use the method of k-nearest neighbors to illustrate the data mining
philosophy of ‘let the data do the talking’.
Date: Wednesday, March 8, 2006
Location: Verizon
Laboratories
Speaker:
Dr. Rama Ramakrishnan,
Title:
Math and the Merchant: The Application of Econometrics, Machine Learning
and Optimization Technologies to Retail Decision-Making
Abstract:
Retailing is a global, multi-trillion dollar industry
and touches just about every person on the planet. Yet, despite its size
and complexity, most retail decisions are made primarily on instinct. This
has started to change in the last decade with the availability of granular
and timely sales data coupled with the development of sophisticated
decision-support software that analyzes this data and recommends optimal
decisions. Retailers are rapidly adopting this technology to help improve
the financial consequences of the numerous decisions they make every day.
In this talk, we focus on decisions made by retailers
in the merchandising process, perhaps the most important business
process within retail. We identify some of the core analytical decision
problems in this area and describe the algorithmic/modeling techniques that
have been developed to address these problems. We highlight the
inter-disciplinary nature of “winning” approaches i.e., they
tend to be hybrids that draw on ideas from disciplines such as
econometrics, machine learning and optimization. We conclude with a summary
of the financial impact of these techniques on retailers’ financials
(and stock prices!) worldwide.
Date: Thursday, January 26, 2006
Location:
Optiant
Speaker:
John Neale
Title:
Supply Chain Inventory Optimization: A Market Description and Case Study
Abstract:
This talk will consist of two parts. The first
part will provide an overview of the supply chain Inventory Optimization
(IO) software market. I will describe the different problems that
constitute the space, the evolution of the market and the different companies
that offer IO solutions, and the operations research challenges of working
in the space. The second part of the talk will present a case study
of a successful IO project at Elmer’s
Products. I will describe Elmer’s
situation and supply chain, the approach taken, the challenges faced (both
technical and organizational), and the results achieved.
Date: Wednesday, December 7, 2005
Location:
Emptoris
Speaker:
Dick Larson
Title:
The OR Profession: Past, Present, Future Prospects and Perils to Avoid
Abstract:
In this talk we discuss
STATE OF THE PROFESSION. How INFORMS is emphasizing
this aspect of Operations Research (OR), using observation, data and
analysis to solve real and important problems. We provide examples of
current 'hot topics' in OR research from many of the different segments of
the profession.
MARKETING THE PROFESSION. Efforts by INFORMS to tell
others about the OR profession, now called "The Science of
Better," including OR's implemented accomplishments and its importance
in efficiently managing organizations.
ARE WE BECOMING TOO NARROW? We will then discuss a
new emerging field in the US
and Europe that increasingly looks like
Phil Morse's OR, while many in our profession are becoming increasingly
narrow, increasingly mathematical.
Date: Wednesday, Oct 19, 2005
Location: Verizon Laboratories
Speaker: Dr. Irv Lustig, Manager of Technical
Services, ILOG Direct
Title: Computational Progess in Optimization
Abstract:
The past 15 years have seen an explosive growth in the
number of applications of linear and integer programming. The evolution of
powerful algorithms for solving these problems--together with ever more
powerful computers--has brought the application of optimization to the
desktop. Irv will describe the computational state-of-the-art in
optimization and the key advances that have led to a tremendous increase in
solving power. The result of this increased power is that many problems
once considered unsolvable are now solvable, often in a matter of minutes.
Other Past Talks:
10/06/04 - Olga Raskina, “Making the
Optimal Better”
05/20/04 - Les Servi, “A Case Study of an Industrial Applied
Probability Research Project"
04/08/04 - Edward Robins, "Evaluating and Maintaining the Value of
'Knowledge' and 'Data' Bases - Methods to Evaluate Current Worth and Worth
Trends"
10/03 - Arnold
Barnett, "Airline Security: Where Are We?"
03/03 - Warren Powell, "Real-Time Optimization for Real-World
Problems"
09/02 - John Hammond, "Making Smart Choices"
Last Update: 2/24/06
|
