Download - Multiplicative Bounds for Metric Labeling

Transcript
Page 1: Multiplicative Bounds for Metric Labeling

Multiplicative Boundsfor Metric Labeling

M. Pawan Kumar

École Centrale Paris

Joint work with Phil Torr, Daphne Koller

Page 2: Multiplicative Bounds for Metric Labeling

Metric Labeling

Variables V = { V1, V2, …, Vn}

Page 3: Multiplicative Bounds for Metric Labeling

Metric Labeling

Variables V = { V1, V2, …, Vn}

Page 4: Multiplicative Bounds for Metric Labeling

Metric Labeling

Va Vb

Labels L = { l1, l2, …, lh}Variables V = { V1, V2, …, Vn}

Labeling f: { 1, 2, …, n} {1, 2, …, h}

E(f) = Σa θa(f(a)) + Σ(a,b) wabd(f(a),f(b))minf

θa(f(a))

θb(f(b))wabd(f(a),f(b))

wab ≥ 0

d is metric

Page 5: Multiplicative Bounds for Metric Labeling

Metric Labeling

Va Vb

E(f)minf

NP hard

= Σa θa(f(a)) + Σ(a,b) wabd(f(a),f(b))

Low-level vision applications

Page 6: Multiplicative Bounds for Metric Labeling

Minka. Expectation Propagation for Approximate Bayesian Inference, UAI, 2001 Murphy et al. Loopy Belief Propagation: An Empirical Study, UAI, 1999 Winn et al. Variational Message Passing, JMLR, 2005 Yedidia et al. Generalized Belief Propagation, NIPS, 2001 Besag. On the Statistical Analysis of Dirty Pictures, JRSS, 1986 Boykov et al. Fast Approximate Energy Minimization via Graph Cuts, PAMI, 2001 Komodakis et al. Fast, Approximately Optimal Solutions for Single and Dynamic MRFs, CVPR, 2007 Lempitsky et al. Fusion Moves for Markov Random Field Optimization, PAMI, 2010 Chekuri et al. Approximation Algorithms for Metric Labeling, SODA, 2001 Goemans et al. Improved Approximate Algorithms for Maximum-Cut, JACM, 1995 Muramatsu et al. A New SOCP Relaxation for Max-Cut, JORJ, 2003 Ravikumar et al. QP Relaxations for Metric Labeling, ICML, 2006 Alahari et al. Dynamic Hybrid Algorithms for MAP Inference, PAMI 2010 Kohli et al. On Partial Optimality in Multilabel MRFs, ICML, 2008 Rother et al. Optimizing Binary MRFs via Extended Roof Duality, CVPR, 2007

.

.

.

Approximate Algorithms

Page 7: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves

Page 8: Multiplicative Bounds for Metric Labeling

Integer Linear Program

Number of facets grows exponentially in problem size

Minimize a linear function over a set of feasible solutions

Indicator xa(i) {0,1} for each variable Va and label li

Page 9: Multiplicative Bounds for Metric Labeling

Linear Programming Relaxation

Schlesinger, 1976; Chekuri et al., 2001; Wainwright et al., 2003

Indicator xa(i) {0,1} for each variable Va and label li

Page 10: Multiplicative Bounds for Metric Labeling

Linear Programming Relaxation

Schlesinger, 1976; Chekuri et al., 2001; Wainwright et al., 2003

Indicator xa(i) [0,1] for each variable Va and label li

Page 11: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves

Page 12: Multiplicative Bounds for Metric Labeling

Move-Making AlgorithmsSpace of All Labelings

f

Page 13: Multiplicative Bounds for Metric Labeling

Expansion AlgorithmVariables take label lα or retain current label

Slide courtesy Pushmeet Kohli Boykov, Veksler and Zabih, 2001

Page 14: Multiplicative Bounds for Metric Labeling

Expansion Algorithm

SkyHouse

TreeGround

Initialize with TreeStatus: Expand GroundExpand HouseExpand Sky

Slide courtesy Pushmeet Kohli

Variables take label lα or retain current label

Boykov, Veksler and Zabih, 2001

Page 15: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves

Page 16: Multiplicative Bounds for Metric Labeling

Multiplicative Boundsf*: Optimal Labeling f: Estimated Labeling

Σa θa(f(a)) + Σ(a,b) sabd(f(a),f(b))

Σa θa(f*(a)) + Σ(a,b) sabd(f*(a),f*(b))

Page 17: Multiplicative Bounds for Metric Labeling

Multiplicative Boundsf*: Optimal Labeling f: Estimated Labeling

≤B

Σa θa(f(a)) + Σ(a,b) sabd(f(a),f(b))

Σa θa(f*(a)) + Σ(a,b) sabd(f*(a),f*(b))

Page 18: Multiplicative Bounds for Metric Labeling

Multiplicative Bounds

Expansion LP

Potts 2 2

Metric 2M O(log h)

TruncatedLinear

2M 2 + √2

TruncatedQuadratic

2M O(√M)

M = ratio of maximum and minimum distance

Page 19: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves– Complete Rounding– Interval Rounding– Hierarchical Rounding

Page 20: Multiplicative Bounds for Metric Labeling

Complete Rounding

Treat xa(i) [0,1] as probability that f(a) = i

Cumulative probability ya(i) = Σj≤i xa(j)

0 ya(1) ya(2) ya(h) = 1

r

Generate a random number r (0,1]

Assign the label next to r

ya(k)ya(i)

Page 21: Multiplicative Bounds for Metric Labeling

Complete Move

Va Vb

θab(i,k) = sabd(i,k) NP-hard

Page 22: Multiplicative Bounds for Metric Labeling

Complete Move

Va Vb

θab(i,k) = sabd’(i,k)

d’(i,k) ≥ d(i,k)

d’ is submodular

Page 23: Multiplicative Bounds for Metric Labeling

Complete Move

Va Vb

θab(i,k) = sabd’(i,k)

d’(i,k) ≥ d(i,k)

d’ is submodular

Page 24: Multiplicative Bounds for Metric Labeling

Complete Move

New problem can be solved using minimum cut

Same multiplicative bound as complete rounding

Multiplicative bound is tight

Page 25: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves– Complete Rounding– Interval Rounding– Hierarchical Rounding

Page 26: Multiplicative Bounds for Metric Labeling

Interval Rounding

Treat xa(i) [0,1] as probability that f(a) = i

Cumulative probability ya(i) = Σj≤i xa(j)

0 ya(1) ya(2) ya(h) = 1ya(k)ya(i)

Choose an interval of length h’

Page 27: Multiplicative Bounds for Metric Labeling

Interval Rounding

Treat xa(i) [0,1] as probability that f(a) = i

Cumulative probability ya(i) = Σj≤i xa(j)

r

Generate a random number r (0,1]

Assign the label next to r if it is within the interval

ya(k)ya(i)

Choose an interval of length h’ REPEAT

Page 28: Multiplicative Bounds for Metric Labeling

Interval Move

Va Vb

θab(i,k) = sabd(i,k)

Choose an interval of length h’

Page 29: Multiplicative Bounds for Metric Labeling

Interval Move

Va Vb

θab(i,k) = sabd(i,k)

Choose an interval of length h’

Add the current labels

Page 30: Multiplicative Bounds for Metric Labeling

Interval Move

Va Vb

θab(i,k) = sabd’(i,k)

Choose an interval of length h’

Add the current labels

d’(i,k) ≥ d(i,k)

d’ is submodular

Solve to update labels

Repeat until convergence

Page 31: Multiplicative Bounds for Metric Labeling

Interval Move

Each problem can be solved using minimum cut

Same multiplicative bound as interval rounding

Multiplicative bound is tight

Kumar and Torr, NIPS 2008

Page 32: Multiplicative Bounds for Metric Labeling

Outline

• Linear Programming Relaxation

• Move-Making Algorithms

• Comparison

• Rounding-based Moves– Complete Rounding– Interval Rounding– Hierarchical Rounding

Page 33: Multiplicative Bounds for Metric Labeling

Hierarchical Rounding

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Hierarchical clustering of labels (e.g. r-HST metrics)

Page 34: Multiplicative Bounds for Metric Labeling

Hierarchical Rounding

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Assign variables to labels L1, L2 or L3

Move down the hierarchy until the leaf level

Page 35: Multiplicative Bounds for Metric Labeling

Hierarchical Rounding

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Assign variables to labels l1, l2 or l3

Page 36: Multiplicative Bounds for Metric Labeling

Hierarchical Rounding

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Assign variables to labels l4, l5 or l6

Page 37: Multiplicative Bounds for Metric Labeling

Hierarchical Rounding

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Assign variables to labels l7, l8 or l9

Page 38: Multiplicative Bounds for Metric Labeling

Hierarchical Move

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Hierarchical clustering of labels (e.g. r-HST metrics)

Page 39: Multiplicative Bounds for Metric Labeling

Hierarchical Move

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Obtain labeling f1 restricted to labels {l1,l2,l3}

Page 40: Multiplicative Bounds for Metric Labeling

Hierarchical Move

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Obtain labeling f2 restricted to labels {l4,l5,l6}

Page 41: Multiplicative Bounds for Metric Labeling

Hierarchical Move

L1 L2

l1 l2 l3 l4 l5 l6 l7 l8 l9

L3

Obtain labeling f3 restricted to labels {l7,l8,l9}

Page 42: Multiplicative Bounds for Metric Labeling

Hierarchical Move

L1 L2 L3

Va Vb

f1(a)

f2(a)

f3(a)

Move up the hierarchy until we reach the root

f1(b)

f2(b)

f3(b)

Page 43: Multiplicative Bounds for Metric Labeling

Hierarchical Move

Each problem can be solved using minimum cut

Same multiplicative bound as hierarchical rounding

Multiplicative bound is tight

Kumar and Koller, UAI 2009

Page 44: Multiplicative Bounds for Metric Labeling

ConclusionMove-Making LP

Potts 2 2

Metric O(log h) O(log h)

TruncatedLinear

2 + √2 2 + √2

TruncatedQuadratic

O(√M) O(√M)

M = ratio of maximum and minimum distance

Page 45: Multiplicative Bounds for Metric Labeling

Open Problems

• Moves for general rounding schemes

• Higher-order energy functions

• Better comparison criterion

Page 46: Multiplicative Bounds for Metric Labeling

Questions?

http://www.centrale-ponts.fr/personnel/pawan