en
0.25
0.5
0.75
1.25
1.5
1.75
2
Challenges for Machine Learning in Computational Sustainability
Published on 2013-01-167499 Views
Research in computational sustainability seeks to develop and apply methods from computer science to the many challenges of managing the earth's ecosystems sustainably. Viewed as a control problem, ec
Related categories
Presentation
Challenges for Machine Learning in Computational Sustainability00:00
Sustainable Management of the Earth’s Ecosystems00:49
Why?01:41
Computer Science can help!02:26
Computational Sustainability02:56
Outline03:41
Example Research Challenges Data Acquisition04:03
Data Interpretation06:03
Data Integration07:25
Model Fitting08:33
State of the Art: STEM Model of Bird Species Distribution10:07
Policy Optimization11:26
State of the Art: Reserve Design from a Species Distribution Model (1)12:24
State of the Art: Reserve Design from a Species Distribution Model (2)12:57
State of the Art: Reserve Design from a Species Distribution Model (3)13:03
Policy Execution13:25
Drill Down: Three Projects at Oregon State15:03
Project eBird www.ebird.org15:52
Species Distribution Modeling from Citizen Science Data17:05
Imperfect Detection18:18
Multiple Visits to the Same Sites19:30
Occupancy-Detection Model20:45
Standard Approach: Log Linear (logistic regression) models22:38
Results on Synthetic Species with Nonlinear Dependencies23:09
A Flexible Predictive (non-Latent) Model24:36
Predictive Model Results25:28
Two Approaches: Summary26:14
The Dream27:32
A Simple Idea: Parameterize 𝐹 and 𝐺 as boosted trees27:46
Results: OD-BRT28:15
Handling Variable Expertise28:31
Expert vs. Novice Differences28:56
BirdCast: Understanding and Forecasting Bird Migration29:48
Modeling Goal: Spatial Hidden Markov Model31:58
Problem: We have only aggregate data32:31
Solution: Collective Graphical Models33:15
Solution: Collective Graphical Models (2)34:20
Inference in Collective Graphical Models35:10
The Migration Model36:03
With Added Covariates36:18
Invasive Species Management in River Networks36:36
Markov Decision Process37:13
Dynamics and Objective38:23
Algorithm DDV (1)39:20
Algorithm DDV (2)40:39
Results on “RiverSwim” benchmark41:12
Published Rule of Thumb Policies for Invasive Species Management41:51
Cost Comparisons: Rule of Thumb Policies vs. DDV42:17
Summary42:26
Distinctive Characteristics of Sustainability Problems43:12
Thank-you45:44
Questions?46:18