Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

10
Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting

TAGS:

Transcript of Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Page 1: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Decision Making

UtilityFunctions

ProbabilisticGraphicalModels

Acting

Page 2: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Utilities and Preferences

0.8

0.2 $4mil

$0 0.75

0.25 $3mil

$0

Page 3: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Utility = Payoff

0.2

0.8 $4mil

$0 0

1 $3mil

$0

Page 4: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

• Fair coin is tossed repeatedly until it comes up heads, say on the nth toss

• Payoff = $2n

St. Petersburg Paradox

Page 5: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

$

U

Typical Utility Curve

Page 6: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

$ reward

U

10000 500

UD

400

Certainty equivalent insurance/risk premium

U($500)

D=$0 with prob. 0.5

$1000 with prob. 0.5

Page 7: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Multi-Attribute Utilities• All attributes affecting preferences

must be integrated into one utility function

• Human life–Micromorts– QALY (quality-adjusted life year)

Page 8: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Testing Knowledge

Loss offetus

Futurepregnancy

Down’ssyndrome

Example: Prenatal diagnosisU1(T) + U2(K) + U3(D,L) + U4(L,F)

Page 9: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

Summary• Our utility function determines our preferences

about decisions that involve uncertainty• Utility generally depends on multiple factors

– Money, time, chances of death, …

• Relationship is usually nonlinear– Shape of utility curve determines attitude to risk

• Multi-attribute utilities can help decompose high-dimensional function into tractable pieces

Page 10: Daphne Koller Decision Making Utility Functions Probabilistic Graphical Models Acting.

Daphne Koller

END END END


Probabilistic Graphical Models · 2014-01-15 · Logistics Text books: Daphne Koller and Nir Friedman, Probabilistic Graphical Models M. I. Jordan, An Introduction to Probabilistic

Probabilistic Graphical Models · 2014-01-15 · Logistics Text books: Daphne Koller and Nir Friedman, Probabilistic Graphical Models M. I. Jordan, An Introduction to Probabilistic

Efficient Solution Algorithms for Factored MDPs by Carlos Guestrin, Daphne Koller, Ronald Parr, Shobha Venkataraman Presented by Arkady Epshteyn.

Efficient Solution Algorithms for Factored MDPs by Carlos Guestrin, Daphne Koller, Ronald Parr, Shobha Venkataraman Presented by Arkady Epshteyn.

Hierarchical Graph Cuts for Semi-Metric Labeling M. Pawan Kumar Joint work with Daphne Koller.

Hierarchical Graph Cuts for Semi-Metric Labeling M. Pawan Kumar Joint work with Daphne Koller.

7 Probabilistic Entity-Relationship Models, PRMs, and ...ai.stanford.edu/~koller/Papers/Heckerman+al:SRL07.pdf · 7 Probabilistic Entity-Relationship Models, PRMs, and Plate Models

7 Probabilistic Entity-Relationship Models, PRMs, and ...ai.stanford.edu/~koller/Papers/Heckerman+al:SRL07.pdf · 7 Probabilistic Entity-Relationship Models, PRMs, and Plate Models

Oregon State University – CS539 PRMs Learning Probabilistic Models of Link Structure Getoor, Friedman, Koller, Taskar.

Oregon State University – CS539 PRMs Learning Probabilistic Models of Link Structure Getoor, Friedman, Koller, Taskar.

Constrained Approximate Maximum Entropy Learning (CAMEL) Varun Ganapathi, David Vickrey, John Duchi, Daphne Koller Stanford University TexPoint fonts used.

Constrained Approximate Maximum Entropy Learning (CAMEL) Varun Ganapathi, David Vickrey, John Duchi, Daphne Koller Stanford University TexPoint fonts used.

Daphne Koller Stanford University Learning to improve our lives.

Daphne Koller Stanford University Learning to improve our lives.

Learning Probabilistic Relational Models Lise Getoor 1, Nir Friedman 2, Daphne Koller 1, and Avi Pfeffer 3 1 Stanford University, 2 Hebrew University,

Learning Probabilistic Relational Models Lise Getoor 1, Nir Friedman 2, Daphne Koller 1, and Avi Pfeffer 3 1 Stanford University, 2 Hebrew University,

© Daphne Koller, 2003 Probabilistic Models of Relational Domains Daphne Koller Stanford University.

© Daphne Koller, 2003 Probabilistic Models of Relational Domains Daphne Koller Stanford University.

Probabilistic Graphical Models Principles and Techniques - Koller, Friedman - Unknown - 2009

Probabilistic Graphical Models Principles and Techniques - Koller, Friedman - Unknown - 2009

Statistical Learning from Relational Data Daphne Koller Stanford University Joint work with many many people.

Statistical Learning from Relational Data Daphne Koller Stanford University Joint work with many many people.

Daphne Koller Message Passing Loopy BP and Message Decoding Probabilistic Graphical Models Inference.

Daphne Koller Message Passing Loopy BP and Message Decoding Probabilistic Graphical Models Inference.

Learning 3: Bayesian Networks Reinforcement Learning Genetic Algorithms based on material from Ray Mooney, Daphne Koller, Kevin Murphy.

Learning 3: Bayesian Networks Reinforcement Learning Genetic Algorithms based on material from Ray Mooney, Daphne Koller, Kevin Murphy.

Daphne Koller,engineer/gedc2013/bio-pdfs/koller.pdf · Daphne Koller is the Rajeev Motwani Professor in the Computer Science Department at Stanford University and the Oswald Villard

Daphne Koller,engineer/gedc2013/bio-pdfs/koller.pdf · Daphne Koller is the Rajeev Motwani Professor in the Computer Science Department at Stanford University and the Oswald Villard

Graphical' Models' Structure' Learning'€¦ · Learning' Probabilis2c' Graphical' Models' BN'Structure' Learning' Daphne Koller Why Structure Learning • To learn model for new

Graphical' Models' Structure' Learning'€¦ · Learning' Probabilis2c' Graphical' Models' BN'Structure' Learning' Daphne Koller Why Structure Learning • To learn model for new

Probabilistic Models of Relational Data Daphne Koller Stanford University Joint work with: Lise Getoor Ming-Fai Wong Eran Segal Avi Pfeffer Pieter Abbeel.

Probabilistic Models of Relational Data Daphne Koller Stanford University Joint work with: Lise Getoor Ming-Fai Wong Eran Segal Avi Pfeffer Pieter Abbeel.

Generalizing Plans to New Environments in Relational MDPs Carlos Guestrin Daphne Koller Chris Gearhart Neal Kanodia Stanford University.

Generalizing Plans to New Environments in Relational MDPs Carlos Guestrin Daphne Koller Chris Gearhart Neal Kanodia Stanford University.

Structured Models for Multi-Agent Interactions Daphne Koller Stanford University Joint work with Brian Milch, U.C. Berkeley.

Structured Models for Multi-Agent Interactions Daphne Koller Stanford University Joint work with Brian Milch, U.C. Berkeley.

Multi-modal robotic perception Stephen Gould, Paul Baumstarck, Morgan Quigley, Andrew Ng, Daphne Koller PAIL, January 2008.

Multi-modal robotic perception Stephen Gould, Paul Baumstarck, Morgan Quigley, Andrew Ng, Daphne Koller PAIL, January 2008.

Multiagent Planning with Factored MDPs Carlos Guestrin Daphne Koller Stanford University Ronald Parr Duke University.

Multiagent Planning with Factored MDPs Carlos Guestrin Daphne Koller Stanford University Ronald Parr Duke University.