Computational Sustainability Carla P. Gomes Institute for - - PowerPoint PPT Presentation
Computational Sustainability Carla P. Gomes Institute for Computational Sustainability Computing and Information Science Cornell University Support : Role of Information Sciences and Engineering in Sustainability Expeditions RISES in
Role of Information Sciences and Engineering in Sustainability RISES DC, Feb. 2011
Expeditions in Computing (CISE) Support:
29 graduate students
24 undergrad. students
Expeditions in Computing (CISE) Support:
ICS members and collaborators for their many contributions towards the development of a vision for research, education, and outreach activities in the new area of Computational Sustainability
3
Gro Brundtland Norwegian Prime Minister Chair of WCED
UN World Commission on Environment and Development,1987.
4
[Nobel Prize with Gore 2007]
+130 countries Global Warming Erosion of Biodiversity
Examples:
ecosystems, 50% of all species of life on earth will be extinct in 100 years.
was 50 years ago and is declining. Worm et al. (2006)
There are no major issues raised in Our Common Future for which the foreseeable trends are favourable.
Wilson, E.O. The Future of Life (2002) .
5
Source: Planetary Boundaries: A Safe Operating Space for Humanity, Nature, 2009
6
Sustainable Development encompasses balancing environmental, economic, and societal needs.
7
Often involving multiple and highly interconnected components and players, in highly dynamic and uncertain environments.
ecosystems.noaa.gov
Smart Grid: Complex Digital Ecosystem Natural Ecosystem
Offer challenges but also opportunities for the advancement of the state of the art in computing and information science.
8
9
10
E.g. Conservation and Biodiversity
I Environment II Socio-Economic Systems
E.g. Poverty mapping and poverty reduction, and harvesting policies.
III Energy
E.g. Smart grid, Material discovery and Biofuels
Farmers Fuel distributors Environmental impact Consumers Energy crops Non-energy crops Social welfare Food supply Gasoline producers
Water quality
Soil quality
Local air pollution
Biodiversity Energy market
Economic impact
11
Ideas from circuit theory for mapping critical linkages in complex landscapes
Brad McRae et al.
Connectivity for mountain in southern California. Blue - low current density (low densities of dispersing mountain lions); Yellow - movement bottlenecks, where connectivity is most vulnerable to hab destruction. Red - high current flow high priority areas for conservation or restoration.
With limited funding and constant threats of habitat loss how do we choose
which habitats to protect so that landscapes will stay well-connected for wild animal species?
Maintaining movement and connectivity across landscapes can help reduce inbreeding, increase genetic diversity, and/or colonize new habitat.
Opportunities for new computational models integrating ecological and economic constraints.
Wildlife Corridors link core biological areas, allowing animal movement between areas. Typically: low budgets to implement corridors.
Goal: preserve grizzly bear populations in the U.S. Northern Rockies by creating wildlife corridors connecting 3 reserves: Yellowstone National Park
Glacier Park / Northern Continental Divide Salmon-Selway Ecosystem
cost suitability
14
Wildlife corridor design
Computational problem " Connection Sub-graph Problem Find a sub-graph of G that: contains the reserves; is fully connected; with cost below a given budget; and with maximum utility Connection Sub-Graph - NP-Hard Given a graph G with a set of reserves:
Connection Sub-graph Problem
Worst Case Result --- Real-world problems possess hidden structure that can be exploited allowing scaling up of solutions.
Conrad, Dilkina, Gomes, van Hoeve, Sabharwal, Sutter; 2007-2010
Interdisciplinary Research Project:
25 km2 hex 1288 Cells $7.3M 2 hrs 50x50 grid 167 Cells $1.3B <1 sec 40x40 grid 242 Cells $891M <1 sec 25x25 grid 570 Cells $449M <1 sec 10x10 grid 3299 Cells $99M 10 mins 25 km2 hex Extend with 2xB=$15M 10x in Util
Need to solve problems large number of cells! ">64*4.+*+58'!,,(&,''
Captures Better the Connectedness Structure #! Exponential Number of Constraints $! Provides good upper bounds (discover good cuts with ML)$! Quite compact (poly size)
Dilkina and Gomes 2010
How is hardness affected as the budget fraction is varied?
Problem evaluated on semi-structured graphs m x m lattice / grid graph with k terminals Inspired by the conservation corridors problem Place a terminal each on top-left and bottom-right Maximizes grid use Place remaining terminals randomly Assign uniform random costs and utilities from {0, 1, !, 10}
Utility Gap (Optimally Extended Min cost/ Optimal) Runtime From 6x6 to 10x10 grid (100 parcels): 1000 instances per data-point; Runtime for Optimal Solution
No reserves:
3 reserves
Scaling up Solutions by Exploiting Structure
Static/Dynamic Pruning
5 km grid (12788 land parcels): minimum cost solution 5 km grid (12788 land parcels): +1% of min. cost
Glacier Park Yellowstone Salmon-Selway Real world instance: Corridor for grizzly bears in the Northern Rockies, connecting: Yellowstone Salmon-Selway Ecosystem Glacier Park (12788 nodes)
Our approach allows us to handle large problems and reduce corridor cost dramatically (hundreds of millions of dollars) compared to existing approaches while providing guarantees of optimality in terms of utility: Optimal or within 1% of optimality for interesting budget levels.
19
Grizzly Bear Wolverines Lynx
Collaborators: Michael K. Schwartz USDA Forest Service, Rocky Mountain Research Station Claire Montgomery Oregon State University
Identification of new problems to address multiple species: E.g. Steiner Multigraph Problem Upgrading Shortest Path Problem Dynamics and Game Theory :
20
Pedagogical Games Shortest path, Steiner trees, and much more about Computational Sustainability Lots of undergrad/Meng students Having fun designing games for A Travel Museum on Computational Sustainability
Edutainment
middle school Boynton Middle School Math Day
Effort led by David Schneider
21
[Faloutsos, McCurley, Tompkins 04] Network of Pandemic Influenza
Robert J. Glass,* Laura M. Glass,† Walter E. Beyeler,* and
Facebook Network If a person is infected with a disease, who else is likely to be?
Sensor Web as predicted by Matt Heavner at University of Alaska Southeast.
Sensor and Wireless Networks
22
Information Sciences
Red Cockaded Woodpecker (RCW) is a federally endangered species Current population is estimated to be about 1% of original stable population (~12,000 birds)
Conservation Funds manages Palmetto Peartree Preserve (North Carolina) 32 active RCW territories (as of Sept 2008)
>?&*2$#@'AB@'A+*C+#4@'DB@'E?/42+F42&?@'GB@'H*/46?5$(.@'EB@'0+#,&5?@'IB@'J$#%42@'KB@' L$/&,@'JB@'>4.?4%M4*@'EB'>?/$8,@'AB@'E/(#2,&#@'NB@'E**&#@'OB@'P4("?4#@'DBQ'RSSTUVS
Goal: Increase RCW population level Management options: Prioritizing land acquisition adjacent to current RCW populations Building artificial cavities Translocation of birds
Computational Challenge: scaling up solutions for considering a large number of years (100+) " " decomposition methods and exploiting structure Red Cockaded W. Biological and Ecological Model Management Decisions: Land acquisition Artificial cavities Translocation of birds Must explicitly consider interactions between biological/ecological patterns and management decisions
i j k pkj
pij
1-! t = 1 t = 2 t = 3
Stochastic diffusion model (movement and survival patterns) in RCW populations Stochastic optimization model Decisions: where and when to acquire land parcels Goal: Maximize expected number of surviving RCW
Maximizing the Spread of Cascades
Approaches: Sample Average Approximation Upper confidence Bounds
Spread of rumor/news/articles on Facebook, Twitter, or among blogs/ websites.
–! In human networks, or between networks of households, schools, major cities, etc. –! In agriculture settings.
Viral Marketing
[Domingos and Richardson, KDD 2001]
[Kempe, Kleinberg, and Tardos, KDD 2003]
[Krause and Guestrin 2007]
26
27
Eastern Phoebe Migration
Steven Philips, Miro Dudik & Rob Schapire
Maxent
Information Sciences
K(#'W(@'O&#"UG&&#'O$#"@'I&.&664'X(56?+#,$#'<RSVS=B'
Kelling Active research area bringing together machine and statistical learning: H15&#,+$#'$:'N66(34#68UA&5&6)$#'/$2&*'$:' Y46C&#F+&'&5'4*B'RSSZ'
28
The Citizen Science project at the Lab of Ornithology at Cornell empowers everyone interested in birds — from research labs to backyards to remote forests, anyone who counts birds – to contribute to research.
Gather meaningful data to answer large-scale
research questions
Enable participants to experience the process of scientific investigation and develop problem- solving skills
Apply results to science-based conservation
efforts Examples of research and outreach outputs:
CLO: Track record
conservation science.
Online Research Kit The State of The Birds
8009000$:&;34&#'$
>3&6+&,'2+,5%+.()$#'Y$2&*,' B'
<3#2$%&.*+$
(,&4"&'
!&2*:$+*'=:(''
&D+%2' !*(*&+&:&>?$ !-@A)$B$ C*6&(*$ '*#'"#>$23(3$
E3#$ )*F$ @*;$ E=#$ GF+$
Information Sciences
eBird
How to reduce uncertainty & increase accuracy of species distribution models
More --- next talk by Bill Tomlinson
!! Automated Land change
Evaluation, Reporting and Tracking System (ALERT)
Planetary Scale Information System for assessment of disturbances in the global forest ecosystem:
deforestation, conversion to agriculture
!! This system will help
quantify the carbon impact of changes in the forests
Understand the relationship to global climate variability and human activity
Provide monitoring and verification system needed for the UN REDD+ program for saving tropical forests
!! Provide ubiquitous web-based
access to changes occurring across the globe, creating public awareness
Slide by Vipin Kumar
Planetary Skin ! a global nervous system that will integrate land-sea-air-and-space-based senso helping the public and private sectors make decisions to prevent and adapt to climate change. Time
Planetary Skin Institute Information technology can make a huge impact on the environmental community by helping develop global datasets (e.g., for water, food, energy), that are key for sustainability research.
aR'
!! Want most useful information at minimum cost !! Exploit structure to efficiently find near-optimal solutions
–! Many important sensing utility functions are submodular "E.g.: Greedy algorithm finds near-optimal set of observations
!! Resulting algorithms perform well on real applications
–! E.g., top score in Battle of the Water Sensor Networks challenge –! Can handle robustness, complex constraints, dynamics, !
Slide by Andreas Krause
33
Neo D. Martinez Pacific Ecoinformatics and Computational Ecology Lab
www.FoodWebs.org
34
Neo D. Martinez Pacific Ecoinformatics and Computational Ecology Lab
www.FoodWebs.org
35
Neo D. Martinez Pacific Ecoinformatics and Computational Ecology Lab
www.FoodWebs.org
–! Models make relatively reliable predictions at global scale for ancillary variables such as
temperature
–! But provide the least reliable predictions for variables that are crucial for impact assessment
such as regional precipitation
Regional hydrology exhibits large variations among major IPCC model projections
Disagreement between IPCC models
Predictive Modeling Enable predictive modeling of
typical and extreme behavior from multivariate spatio-temporal data
Relationship Mining
Enable discovery of complex dependence structures: non- linear associations or long range spatial dependencies
Complex Networks
Enable studying of collective behavior of interacting eco-climate systems
High Performance Computing
Enable efficient large-scale spatio- temporal analytics on exascale HPC platforms with complex memory hierarchies
Transformative Computer Science Research
–! Data-guided uncertainty reduction by blending physics models and
data analytics
–! A new understanding of the complex nature of the Earth system and
mechanisms contributing to adverse consequences of climate change
–! Inclusion of data-driven analysis as a standard part of climate
projections and impact assessment (e.g., for IPCC)
... data-intensive science [is] …a new, fourth paradigm for scientific exploration." - Jim Gray
Project aim: A new and transformative data-driven approach that complements physics-based models and improves prediction of the potential impacts of climate change
Lead PI: Vipin Kumar
37
H&.*+(?$63F'I$ U!'\,&'/(*)3*&'2454',&5,'5$'&,)/45&'4#2'/43' 3$9&%58'34[&%#,'#$5'2+%&65*8'/&4,(%&2B' U!'Y46?+#&'*&4%#+#"'4#2'%&*45&2'/&5?$2,'64#' 3&%/+5'/$%&'&b6+'(,&'$:'2454':%$/'94%+&2' ,$(%6&,'"'4#4*$"8'5$',3&6+&,'2+,5%+.()$#'.(5' *+[*&'%&,&4%6?'?4,'.&&#'2$#&B' H14/3*&_'RSSR'\"4#24'3$9&%58'/43'' ' J?%+,'D4%%&[''4#2'J$%&8'c4#"'
,3+>*4#>$63F'I$ $ ''How to estimate the impact and marginal returns of different assets?
Poverty interventions need to be targeted to specific areas Asset-based investments have spatially-varying marginal returns
' Policies for Poverty Reduction: Which set of interventions to apply to each targeted area to maximize the poverty reduction, subject to resource constraints (e.g. budget))? Challenging problems at the intersection of machine learning and optimization, applied economics and social science. J?%+,'D4%%&['4#2''J$%&8'c4#"'
Average marginal returns of different assets (Uganda)
40
Catastrophe Modeling for Rwandan Disease Surveillance Can mobile phones be used as an early warning system for disease outbreaks? Bayesian anomaly detection algorithms
Spatiotemporal Diffusion of Contraceptive Norms How do contraceptive norms spread through rural areas of the developing world? Spatiotemporal diffusion models have the potential to better evaluate the efficacy of HIV prevention techniques and inform policy decisions related to public health. –
Many other exciting projects!!!!
41
!! The biomass of fish is estimated to be 1/10 of what it was 50 years ago and is declining (Worm et al. 2006).
There is therefore a clear urgency of finding ways of defining policies for managing fisheries is a sustainable manner. Over Harvesting
43
Economy Increasing Complexity: more complex models and multiple species interactions Example of a Biological Growth Function F(x): Logistic map: x t+1 = r xt (1 - xt), r is the growth rate
We are interested in identifying policy decisions (e.g. when to open/close a fishery ground over time).
Combinatorial optimization problems with an underlying dynamical model.
Class of Computationally Hard Hybrid Dynamic Optimization Models
Clark 1976; Conrad 1999; Ang, Conrad, Just, 2009; Ermon , Conrad, Gomes and Selman 2010
Pacific Halibut Fishery in Area 3A
Ermon , Conrad, Gomes and Selman UAI 2010
E/$(#5'*&d' e&M',5$6C',+F&' >5$6?4,)6+58' Goal: policy for optimal exploitation
" natural MDP formulation Resource growing over time Uncertain dynamics Economics of harvesting Ermon , Conrad, Gomes and Selman UAI 2010
,' >' ?4%9&,5' "%$M5?'
5$'>'
"%$M' '
Optimal policies for management of renewable resources using MDPs Total Allowable Catch Policy " Constant Proportion Policy "rate of fishing ~ 12.3% of estimated stock. Our Result: Policies with periodic closures of a fishery can outperform Constant Proportion Policies Ermon, Conrad, Gomes, Selman, UAI 2010
General Approach to model renewable resources with complex dynamics!!!! Optimal policies for management of renewable resources using MDPs Ermon, Conrad, Gomes, Selman, UAI 2010
49
–! For example, in the United States buildings consume 39% of the energy used nationwide and 70% of the electric energy used.
Smart buildings
51
But increasing energy efficiency is not the main path towards reducing carbon emissions
52
Renewable Energy
53
For example consumers will have smart meters that can 1.! track energy consumption, 2.! monitor individual power circuits in the home, 3.! control smart appliances and actively manage their energy use; 4.! dynamically select different providers, 5.! sell excess energy (e.g., car battery, wind, biomass) to grid, etc
How to integrate the different sources into the smart grid ? challenging dynamical problems!!! What is the best mix of energy generation technologies? What is the best mix of storage technologies? Where to locate renewable energy plants?
US Power Grid
Total energy produced must be equal to the amount consumed by the system loads AT ALL TIMES:
Managing the Power Grid is a Balancing Act!!!
55
Ambitious mandatory goal of 36 billion gallons of renewable fuels by 2022
(five-fold increase from 2007 level)
(Signed into law in Dec. 2007)
Farmers Fuel distributors Environmental impact Consumers Energy crops Non-energy crops Social welfare Food supply Gasoline producers
Water quality
Soil quality
Local air pollution
Biodiversity Energy market
Economic impact
Advanced Biofuels ( Cleaner) (non food crops and biomass )
Switchgrass Wood Waste Animal Waste Municipal Waste
First generation Biofuels
Computational challenges: Large Scale Logistics Planning Realistic Computational Models to Evaluate Impact
56
Feedstock Map
Large-scale investment in new technology provides exciting logistical planning and optimization challenges and opportunities
Biomass Map Potential biorefinery locations Transportation Network (Roads, Rail, Marine) Distribution Terminals and Inter-Modal Facilities to Transfer Liquid Fuel
Resulting optimization models are beyond the scope of the current state of the art in several ways:
and demand) " new models to capture uncertainty "!new stochastic
and capacity
Farmers Fuel distributors Environmental impact Consumers Energy crops Non-energy crops Social welfare Food supply Gasoline producers Water quality Soil quality Local air pollution Biodiversity Energy market Economic impact
Current approaches limited in scope and complexity !! E.g. based on general equilibrium models (e.g., Nash style) !! Strong convexity assumptions to keep the model simple enough for analytical, closed-form solutions (unrealistic scenarios) " Limited computational thinking Transformative research directions !! More realistic computational models in which meaningful solutions can be computed !! Large-scale data, beyond state-of-the-art CS techniques !! Study of dynamics of reaching equilibrium — key for adaptive policy making! Impact on Land-use Impact on food prices?
How to measure risks/ predict rare events? Life Cycle Analysis General Equilibrium Models Models and Metrics!!!!!
Complex Adaptive Systems and Multi-Agent Systems
deling and control of complex high-dimensional systems
machine learning models built from large-scale heterogeneous data.
Farmers Fuel distributors Environmental impact Consumers Energy crops Non-energy crops Social welfare Food supply Gasoline producers Water quality Soil quality Local air pollution Biodiversity Energy market Economic impact
Current approaches limited in scope and complexity !! E.g. based on general equilibrium models (e.g., Nash style) !! Strong convexity assumptions to keep the model simple enough for analytical, closed-form solutions (unrealistic scenarios) " Limited computational thinking Transformative research directions !! More realistic computational models in which meaningful solutions can be computed !! Large-scale data, beyond state-of-the-art CS techniques !! Study of dynamics of reaching equilibrium — key for adaptive policy making! Impact on Land-use Impact on food prices?
How to measure risks/ predict rare events? Life Cycle Analysis General Equilibrium Models Models and Metrics!!!!!
Complex Adaptive Systems and Multi-Agent Systems
Modeling and control of complex high-dimensional systems by combining physics-based models, model-based reasoning,
machine learning models built from large-scale heterogeneous data.
59
60
Agriculture is the primary driver of land use change and deforestation and an important source of Greenhouse Gases (GHG)s: 52% and 84% of global anthropogenic CH4 and N2O, significant C02 Nitrogen Based Fertilizers 31% Source of Greenhouse Gases
Nitrogen Based Fertilizers
61
Spatial and temporal analysis of Nitrogen cycle Study of fertilizers and design of Experiments Collaborator Harold van Es Dead Zone in Gulf of Mexico Fertilizers have played a key role in the increase of food production but they key culprits in the production of greenhouse gases emissions creating dead zones Townsend & Howarth (2010) Scientific American
Perfectly balanced 2.33 Below average Above average
Latin Square Spatially Balanced Latin Square
this is not a completion problem)
63
Streamlining Constraint Reasoning
Discovery of structural properties across solutions (machine learning); Divide (streamline) the search space by imposing such additional properties.
Design of Scientific Experiments for Studying Fertilizers – Spatially Balanced Latin Squares Existence of SBLS – open problem in combinatorics
"!Streamlining "!Scaling up of solutions
Gomes and Sellmann 2003/2004
YES: XOR-Streamlining based on random parity constraints
(Valiant and Vazirani 1986, Unique SAT)
Provable bounds for Counting and Sampling.
Gomes, Sabharwal, Selman, 2006
Discovery of Construction for SBLS for experimental design
LeBras, Perrault, Gomes, 2010
Streamlining by Permutation of Complete Columns of a Cyclic Latin Square
SBLS Order 35
Can we find an domain independent way of streamlining?
64
65
–! Find new products –! Find product substitutes –! Understand material properties
"Sputter 2 or 3 metals (or oxides) onto a silicon wafer (which produces a thin-film) "!Use x-ray diffraction to study crystallographic structure of the thin-film "!Note: electromagnetic radiation experiments are very expensive!! Example: study of platinum-tantulum library showed correlation between crystallographic phase and improved fuel cell
Fuel Cell Applications Ronan LeBras, Damoulas, John M. Gregoire, Sabharwal, Gomes, van Dover
66
67
Fe Al Si
68
Fe Al Si
(38% Fe, 45% Al, 17% Si)
69
Fe Al Si
(38% Fe, 45% Al, 17% Si)
[Source: Pyrotope, Sebastien Merkel]
70
Fe Al Si
(38% Fe, 45% Al, 17% Si)
15 20 25 30 35 40 45 50 55 60 20 40 60 80 100 120
71
Fe Al Si
15 20 25 30 35 40 45 50 55 60 20 40 60 80 100 120 15 20 25 30 35 40 45 50 55 60 20 40 60 80 100 120INPUT:
Fe Al Si
15 20 25 30 35 40 45 50 55 60 20 40 60 80 100 120 15 20 25 30 35 40 45 50 55 60 20 40 60 80 100 120 15 20 25 30 35 40 45 50 55 60 65 10 20 30 40 50 60 70pure phase regions
UNDERLYING PHYSICAL STRUCTURE:
Fe Al Si
Labeled regions mixed phase region mixed phase region
OUTPUT:
NP-hard
72
Two global alignment kernels
Optimization Model
73
Similarity Kernels
for a global data-driven view
A language for
enforcing
Significantly outperforming previous methods.
74
!! A major part of the effort has been identifying a variety of collaborative projects that bring computer scientists together with domain experts in a number of fields !! Individual projects organized under the umbrella of IRPs (Interdisciplinary Research Projects)
–! Currently 13 Application IRPs, 5 Technical IRPs –! Many more in incubation!
Application IRP: Bird Conservation Application IRP: Biofuels Technical IRP: Dynamical Systems Technical IRP: Optimization Species distribution modeling Bird migration RCW preservation Reducing sampling bias Equilibrium models Impact of gasoline taxes Carbon offsets in cap-and-trade
Problem structure Stochastic
Multi-agent inference Balanced designs Parameter estimation Synchronization Monitoring of sustainability Critical transitions Examples:
IRP Name
Researchers Collaborators Students
1.
Reserve Selection and Wildlife Corridors
Conrad, Gomes, Sabharwal, Selman, Shmoys Dilkina, Dimiduk, Drusvyatskiy 2.
Bird Conservation
Lab of O: Rosenberg, Webb TCF: Allen, Amundsen, Vaughan Cornell: Conrad, Damoulas, Gomes, Sabharwal, Shmoys OSU: Dietterich, Hutchinson, Sheldon, Wong Lab of O: Farnsworth, Fink, Hames, Hochachka, Kelling, Wood Cornell: Hooker, Joachims, Riedewald Ahmadizadeh, Castorena, Dilkina, Elmachtoub, Finseth 3.
Fishery Management and Fish Preservation
Conrad, Gomes, Sabharwal, Selman, Yakubu, Zeeman Jerald, Ziyadi Alam, Ermon, Haycraft, Li, Wiley 4.
Disease Prevention in Dairy Cows
Damoulas, Ellner, Gomes, Sabharwal, Selman Grohn, Cha Ahmadizadeh, LeBras, Prakash, Smith, Teose 5.
Forest Fires
Borradaile, Dietterich, Montgomery, Rusmevichientong, Shmoys, Wong Ferst, Spencer, Gagnon, Houtman, Mark, Reynolds
continued !
IRP Name
Researchers Collaborators Students
6.
Modeling Ecological Dynamics
Ellner Jones, Mao-Jones, Ritchie, King, Rohani, Ionides, Kendall, Reumann, Newman, Ferrari, Hooker 7.
Biofuels
Bento, Walker Anderson Bergstrom, Cuvilliez, Kang, Klotz, Landry, Latza, Mayton, Patel, Roth, Stanislaus, Taranto, Weiz 8.
Pastoral Systems in East Africa
Barrett, Gomes, Selman, Wong Kariuki, Leeuw, Smith Guo, Toth, Zhuo 9.
Poverty Mitigation and Food Insecurity
Barrett, Damoulas, Gomes, Lentz, Naschold Allen, Bell, Lang, Maxwell, McGlinchy, Murray Dilkina, Guo, Harou, Lang 10.
Invasive Species
Conrad, Shmoys Dietterich,Gomes Hutchinson, Sheldon Yang, Ermon, Spencer 11.
Impact of Climate Change
Guckenheimer, Mahowald, Zeeman Many colleagues in climate science 12.
Stress/Environment Interactions
Zeeman McCobb Gribizis 13.
Material Discovery
Damoulas, Gomes, Sabharwal Gregoire, van Dover LeBras
78
IRP Name
Researchers Collaborators Students
14.
Dynamical Systems
Guckenheimer, Strogatz, Zeeman Haiduc, Kuehn 15.
Network Science
Gomes, Hopcroft, Selman, Strogatz Ermon, Sheldon 16.
Bridging Constraint Reasoning and Machine Learning
Damoulas, Dietterich, Gomes, Sabharwal Gregoire, van Dover LeBras 17.
Combinatorial Reasoning and Stochastic Optimization
Gomes, Sabharwal, Selman, Shmoys, Sheldon Malitski, Sellmann Cheung, Dilkina, Ermon, Guo, Kroc, LeBras, Perrault, Ramanujan 18.
Large Scale Computing Paradigms for Massively Large Data Sets and Simulation-Intensive Studies
Gomes, Sabharwal, Selman Ahmadizadeh, Dilkina, Kroc
79
Constraint Satisfaction and Optimization Data and Uncertainty Distributed Highly Interconnected Components / Agents Complexity levels in Computational Sustainability Problems Dynamics
Study computational problems as natural phenomena " Science of Computation Many highly interconnected components; " " From Centralized to Distributed: multi agent systems Complex dynamics and Multiple scales " " From Statics to Dynamics: Complex systems and dynamical systems Large-scale data and uncertainty " " Machine Learning, Statistical Modeling
Design of policies to manage natural resources translating into large-scale optimization and learning problems, combining a mixture of discrete and continuous effects, in a highly dynamic and uncertain environment " increasing
levels of complexity
Complex decision and optimization problems " Constraint Reasoning and Optimization
80
Data & Machine Learning
DietterichPI WongCo-PI ChavarriaH
Environmental & Socioeconomic Needs Dynamical Models
Guckenheimer Strogatz ZeemanPI,W YakubuAA
Constraint Reasoning & Optimization
Gomes PI,W
, Hopcroft Co-PI
SelmanCo-PI
, ShmoysCo-PI
Resource Economics,
Environmental Sciences & Engineering
AlbersCo-PI,W, Amundsen, Barrett, Bento, ConradCo-PI, DiSalvo, MahowaldW, MontgomeryCo-PI,W Rosenberg,SofiaW, WalkerAA
Transformative Synthesis: Dynamics & Learning
Science of Computation Science of Computation
Study computational problems as natural phenomena " Science of Computation Many highly interconnected components; " " From Centralized to Distributed: multi agent systems Complex dynamics and Multiple time scales " " From Statics to Dynamics: Complex systems and dynamical systems Large-scale data and uncertainty " " Machine Learning, Statistical Modeling Complex decision and optimization problems " Constraint Reasoning and Optimization
Design of policies to manage natural resources translating into large-scale optimization and learning problems, combining a mixture of discrete and continuous effects, in a highly dynamic and uncertain environment
" increasing levels of complexity
ICS
Coordinating transformative synthesis collaborations Interdisciplinary Research Projects (IRPs) Web Portal Seminars
Host visiting Scientists Conference & Workshops External Collaborations Summer REU program targeting minority students Computational Sustainability courses Research seminar series Postdocs Doctoral students Honors projects
Cornell Cooperative Extension Citizen Science
Conservation Fund Rocky Mountain Research Station The Nature Conservancy K-12 Activities Travel Museum
CompSust09: Cornell, 2009 Over 225 international researchers from several
disciplines and institutions (universities, labs, government)
International Workshop on Constraint Reasoning and Optimization for Computational Sustainability
NIPS-09 Mini Symposium: Machine Learning for Sustainability
Blogs:
CompSust 10: MIT, 2010 AAAI 2011 Special Track on Computational Sustainability February 3, 2011: abstracts due February 8, 2011: papers due
84
85
Sustainability
Climate Natural Resources Social Factors Human Factors Education Economic factors Energy Public Policy
86
Key challenge: to effectively and efficiently establish interdisciplinary collaborations – the level of interconnectedness of social, economic, and environmental issues makes it really challenging!
87
New challenging applications for Computer Science Computational Thinking that will provide new insights into sustainability problems:
New methodologies In Computer Science New methodologies In Sustainability fields
88
89
From the Cloud Institute, NY
Computational Sustainability Statistics eScience
Networks & Systems
HCI
Game Theory
Operations Research
Systems Eng.
Appl. Mathematics
AI
Alg. Theory
Climate Natural Resources Social Factors Human Factors Education Economic factors Energy Public Policy
90
91
Table of Product Characteristics id Property name Value MilkProd productsrep MilkA MilkProd quantity 10000 MilkProd validity date 10/06/2006 CheeseProd products rep Minas CheeseProd quantity 2000 CheeseProd validity date 12/02/2006 CheeseProd shape Circular
Deluge of heterogeneous data from distributed noisy sensors: Need for Technical skills Socio-emotional skills Transcultural skills
–! Use social science to understand consumer attitudes
–! Make sustainability economically relevant
–! Perception as Environmental Leader –! Energy reductions in operations and products –! New markets, products, services
–! Facilitate ubiquitous monitoring-inferencing- actuation –! Simplify device and data management for mere mortals
93
Source : EPRI Source : ASU
Slide by Eve M. Schooler
Existing and proposed wind farms in US and Mexico (2008)
Build-out to reach potential would require >1.7 million turbines
also often associated with migratory pathways Andrew Farnsworth, Ken Rosenberg, 2009
IWhere to not locate wind farms? Guidelines for locating wind farms
Ken Rosenberg (ICS) part of the science team for the State of the Birds Report