Planning & Management

Oral

394904 - Adaptive Human-guided Search Operators for Interactive Optimization of Watershed Conservation Plans

Thursday, June 7
10:30 AM - 12:00 PM
Location: Greenway GH
Co-Authors: Meghna Babbar-Sebens, Oregon – Oregon State University

Interactive Genetic Algorithm (IGA), a type of interactive optimization technique, enables users to define new constraints and/or objectives that reflect their subjective preferences. These user-defined objective functions may be based on Likert-scale type user ratings, and are directly included as additional objective functions within the algorithm’s search process. For obtaining user-provided values of user ratings, the algorithm requires interaction with the user via a graphical user interface, every iteration of the search. While previous research has shown promise in the technique’s ability to search for user-preferred design alternatives, these algorithms can suffer from poor convergence and human fatigue, especially if the decision space is large. In conservation planning problems, users may exhibit preferences for specific decisions (e.g., user may prefer cover crops over wetlands on their field) in addition to (or, in some cases, instead of) preferences for physical objective functions (e.g., reduction in peak flows). Hence, in this work, we examine if emerging patterns in user-preferred decisions during the search process can be used to adapt the IGA’s search operators. New Adaptive Human-guided Search (AHS) operators that incorporate these learned patterns in decision space were proposed. To evaluate their effectiveness, we generated six simulated users (three deterministic users and three stochastic users), based on user-interaction data of three real participants. Convergence was measured using the percent of design alternatives in the Genetic Algorithm population that had user rating classified as “I like it” by the participant. Results show a faster converge rate for deterministic and stochastic participants.

Adriana Debora. Piemonti

Water Resorces Engineer
INTERA, inc

Presentation(s):

Send Email for Adriana Piemonti


Assets

394904 - Adaptive Human-guided Search Operators for Interactive Optimization of Watershed Conservation Plans



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Adaptive Human-guided Search Operators for Interactive Optimization of Watershed Conservation Plans