|
Activity Number:
|
177
- Observations Through a Foggy Lens: Modeling Complex Measurement Error and Non-Random Missingness in Ecological and Environmental Health Data
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 8, 2022 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Consulting
|
|
Abstract #320606
|
|
|
Title:
|
Reinforcement Learning and Step Selection Analysis for Animal Movement Data
|
|
Author(s):
|
Toryn L J Schafer* and Christopher K. Wikle and Mevin B Hooten
|
|
Companies:
|
Cornell University and University of Missouri and University of Texas at Austin
|
|
Keywords:
|
reinforcement learning;
animal behavior;
bayesian estimation;
ecology
|
|
Abstract:
|
Animal movement models are inferential tools for data collected by telemetry devices on animals. One common class of models, the step-selection analysis, is closely related to a model emerging from optimal control literature, the linearly-solvable Markov decision process (LMDP). We compare and contrast the two modeling frameworks in order to demonstrate the interpretation of animals on the landscape as active learners. Simulations illustrate trial and error learning by agents (animals) with memory and complex interactions between the environment and agents. Inference under the LMDP is done in a Bayesian framework with approximate inference obtained by variational inference in large state space settings.
|
Authors who are presenting talks have a * after their name.
