Systems Science Friday Noon Seminar Series

Download

Download (1.7 MB)

Date

12-4-2009

Abstract

Being told to give two different, and potentially counter, responses to the same stimulus can set up a double bind in humans, leading to a type of insanity. So what how do you deal with it when it comes up quite frequently in modeling through simplification and removal of predictive variables?

In his current dissertation research Ike Eisenhauer is using reconstructability analysis to implement K-System, U-System, and B-System approaches to predict a continuously valued function through discrete categorically valued input variables [e.g. textual data]. One of the key issues is how to address the inability of K-Systems and U-Systems to allow the same input to give two different outputs, as well as how to report the performance of learning predictive systems which are trained to know that the multiplicity will exist. This discussion session will consist of a quick overview of Ike's current work and then discuss the key point of: If a system has learned [been trained] that there are two or more different "correct" responses to a given stimulus, what should it report if it is only allowed to pick one response? Especially when it is "punished strongly" for not giving the other one, regardless.

Biographical Information

William "Ike" Eisenhauer is an adjunct assistant professor in both the Systems Engineering and Engineering and Technology Management departments at Portland State University. His research interests include: Adaptive Belief Management, State Based Reconstructability Analysis, Shared Resource Constrained Data Envelope Analysis, Conflict Under Deceptive Irrationality, and Sustainable Quality Management Program Development. In addition to teaching, Ike is Chief of Systems Engineering for the U.S. Department of Veterans Affairs Portland VA Medical Center. His work there is focused on the systemic improvement of health care delivery. Prior to joining the DVA, Ike held positions at Wells Fargo in Risk and Loss Management and Equity Operations. He is an industry consultant in the areas of probability/uncertainty management, executive decision making, and benchmarking. His past clients have included US Bank, Hollywood Entertainment, and Multnomah County.

Subjects

Reconstructability analysis, Predictive control, Text data mining, Double bind (Psychology), Machine learning

Disciplines

Other Engineering | Theory, Knowledge and Science

Persistent Identifier

https://archives.pdx.edu/ds/psu/31068

Rights

© Copyright the author(s)

IN COPYRIGHT:
http://rightsstatements.org/vocab/InC/1.0/
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

DISCLAIMER:
The purpose of this statement is to help the public understand how this Item may be used. When there is a (non-standard) License or contract that governs re-use of the associated Item, this statement only summarizes the effects of some of its terms. It is not a License, and should not be used to license your Work. To license your own Work, use a License offered at https://creativecommons.org/

Creating Insanity in Learning Systems: Addressing Ambiguity Effects of Predicting Non-linear Continuous Valued Functions with Reconstructabilty Analysis from Large Categorically Valued Input Data Sets

Share

COinS