Computations in Science Seminars
Jul
30
Wed 12:15
Brian Skinner, Argonne
e-mail:
Host: Leo Kadanoff ()
Problems in human motion planning

Moving through a densely-populated environment can be surprisingly hard, owing to the problem of congestion. Learning to deal with congestion in crowds and in networks is a long-standing and urgently-studied problem, one that can be equally well described at the level of dense, correlated matter or at the level of game-theoretical decision making.

In this talk I describe two related problems associated with human motion planning. In the first part I consider a description of pedestrian crowds as densely-packed repulsive particles, and I address the question: what is the form of the pedestrian-pedestrian interaction law? Starting with real-life crowd data, I show that pedestrian interactions are described by a remarkably simple power law. Unusually, this interaction is based not on the physical separation between pedestrians but on the imminence in time of a potential future collision.

In the second part of the talk I examine a simple model of a traffic network and study how inefficiency in the traffic flow arises from "selfish" decision-making. I show that, for networks comprised of fast roads and slower roads, the network flow becomes maximally inefficient precisely when the proportion of fast roads matches the network percolation threshold. This conclusion suggests a surprising connection between Nash equilibria from game theory and percolative phase transitions from statistical physics.

Aug
6
Wed 12:15
Andrew Gronewold, Great Lakes Environmental Research Laboratory
e-mail:
Host: Leo Kadanoff ()
Drivers of water level change on Earth’s largest lake system

In December 2012 and January 2013, water levels on Lake Michigan-Huron (the single largest area of fresh surface water on Earth) dropped to record lows (based on a record dating to the mid-1800s). This hydrological event occurred during a period (that begin in the late 1990s) in which the North American Great Lakes have been characterized by above average water temperatures, high evaporation rates, and persistent low water levels. Recent research suggests, however, that the extreme cold winter of 2013-2014 may have significantly lowered the heat content of the Great Lakes, and could signify a transition between hydrological and thermal regimes.

Here, I explore historical drivers of water level change across the Great Lakes, with a particular emphasis on analyzing model simulations and forecasts that propagate changes in regional precipitation, temperature, and evaporation into seasonal and interannual water budget and level dynamics. I assess the skill of these models, and underscore periods when they have performed well, and when they have failed to adequately explain observed water level variability. I conclude with a discussion of water level projections, and the scientific research needed to improve their skill over different time scales.

Aug
13
Wed 12:15 PM
William A. Dembski, Discovery Institue
e-mail:
Conservation of Information in Evolutionary Search

Conservation of Information (CoI) asserts that the amount of information a search outputs can equal but never exceed the amount of information it inputs. Mathematically, CoI sets limits on the information cost incurred when the probability of success of a targeted search gets raised from p to q (p < q), that cost being calculated in terms of the probability p/q. CoI builds on the No Free Lunch (NFL) theorems, which showed that average performance of any search is no better than blind search. CoI shows that when, for a given problem, a search outperforms blind search, it does so by incorporating an amount of information determined by the increase in probability with which the search outperforms blind search. CoI applies to evolutionary search, showing that natural selection cannot create the information that enables evolution to be successful, but at best redistributes already existing information. CoI has implications for teleology in nature, consistent with natural teleological laws mooted in Thomas Nagel's Mind & Cosmos.

Aug
20
Wed 12:15
Glen Weyl, University of Chicago
e-mail:
Host: Leo Kadanoff ()
"Quadratic Voting"

Democratic government is famously plagued by problems such as the tyranny of the majority and political paralysis. These result from fundamental flaws of one-man-one-vote as a mechanism for collective decisions identified by economists. A simple alternative procedure, Quadratic Voting (QV), solves these problems and offers a practical, efficient, simple and robust alternative. Individuals purchase votes using either money or an artificial currency that may be spread across multiple issues and pay the square of the number of votes purchased. QV is efficient because individuals optimally equate the marginal cost of a vote to the benefit they derive from an additional vote and thus set the number of votes purchased proportional to their values for the outcome if all individuals take the chance of their being pivotal in the outcome as approximately constant. This "should" be true in large populations and a number of detailed approximate calculations, intermediate analytic results and numerical simulations have persuaded us of this. However, we have as of yet been unable to rigorously prove the convergence results that we conjecture in detail due to the small probability of a single individual purchasing a large number of votes that turns out to be necessary to sustain equilibrium. After a brief introduction to the motivation, I plan to devote most of the talk to these formal difficulties in hopes of soliciting suggestions on how to clear these roadblocks.

Aug
27
Wed 12:15
Sayantan Majumdar, University of Chicago
e-mail:
Host: Leo Kadanoff ()
Oct
1
Wed 12:15
Stas Nagy, University of Chicago
Host: Leo Kadanoff ()
snagy@uchicago.edu
Oct
8
Wed 12:15
Bob Batterman, University of Pittsburgh
e-mail:
Host: Leo Kadanoff ()
Oct
15
Wed 12:15
Jane Wang, Cornell
e-mail:
Host: Leo Kadanoff ()
Oct
22
Wed 12:15
Andy Ruina, Cornell
e-mail:
Host: Leo Kadanoff ()
Oct
29
Wed 12:15
Kerry Emanuel, MIT
e-mail:
Host: Leo Kadanoff ()
Nov
5
Wed 12:15
Guenter Ahlers, UC Santa Barbara
e-mail:
Host: Leo Kadanoff ()
Nov
12
Wed 12:15
Luis Bettencourt, Santa Fe Institute
e-mail:
Host: Leo Kadanoff ()
Nov
19
Wed 12:15
Emil Martinec, University of Chicago
e-mail:
Dec
3
Wed 12:15
Susan Coppersmith, University of Wisconsin
e-mail:
Host: Leo Kadanoff ()
Dec
10
Wed 12:15
Igor Aronson, Argonne
e-mail:
Host: Leo Kadanoff ()
Jan 2015
7
Wed 12:15
OPEN
Jan 2015
14
Wed 12:15
OPEN
Jan 2015
21
Wed 12:15
Heinrich Jaeger, University of Chicago
e-mail:
Host: Leo Kadanoff ()
Jan 2015
28
Wed 12:15
Seth Lloyd, MIT
e-mail:
Host: Leo Kadanoff ()
Feb 2015
4
Wed 12:15
OPEN
Feb 2015
11
Wed 12:15
OPEN
Feb 2015
18
Wed 12:15
OPEN
Feb 2015
25
Wed 12:15
OPEN
Mar 2015
4
Wed 12:15
OPEN
Mar 2015
11
Wed 12:15
OPEN
Mar 2015
18
Wed 12:15
OPEN
Mar 2015
25
Wed 12:15
OPEN
Apr 2015
1
Wed 12:15
Andrea Bertozzi, UCLA
e-mail:
Host: Leo Kadanoff ()