BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:0
UID:UW-Physics-Event-2644
DTSTART:20120424T170500Z
DURATION:PT1H0M0S
DTSTAMP:20210620T045622Z
LAST-MODIFIED:20120406T143748Z
LOCATION:4274 Chamberlin
SUMMARY:Adaptive information\, Chaos & Complex Systems Seminar\, Rob N
owak\, UW Department of Electrical and Computer Engineering
DESCRIPTION:This talk will discuss the notions of adaptive and non-ada
ptive information in the context of statistical learning and inference
. Suppose that we have a collection of models (e.g.\, signals\, syste
ms\, representations\, etc.) denoted by X and a collection of measurem
ent actions (e.g.\, samples\, probes\, queries\, experiments\, etc.) d
enoted by Y. A particular model x in X best describes the problem at h
and and is measured as follows. Each measurement action\, y in Y\, ge
nerates an observation y(x) that is a function of the unknown model.
This function may be deterministic or stochastic. The goal is to iden
tify x from a set of measurements y_1(x)\,...\,y_n(x)\, where y_i in Y
\, i=1\,...\,n. If the measurement actions y_1\,...\,y_n are chosen d
eterministically or randomly without knowledge of x\, then the measure
ment process is non-adaptive. However\, If y_i is selected in a way th
at depends on the previous measurements y_1(x)\,...\,y_{i-1}(x)\, the
n the process is adaptive. Adaptive information is clearly more flexib
le\, since the process can always disregard previously collected data.
The advantage of adaptive information is that it can sequentially fo
cus measurements or sensing actions to distinguish the elements of X t
hat are most consistent with previously collected data\, and this can
lead to significantly more reliable decisions. The idea of adaptive i
nformation gathering is commonplace (e.g.\, humans and animals excel a
t this)\, but outside of simple parametric settings little is known ab
out the fundamental limits and capabilities of such systems. Some pr
eliminary results addressing this situation will be described.
URL:https://www.physics.wisc.edu/events/?id=2644
END:VEVENT
END:VCALENDAR