Handling Outliers and Missing Data in Statistical Data Models

Handling Outliers and Missing Data in Statistical Data Models

Kaushik MitraDate: 17/1/2011ECSU Seminar, ISI

Statistical Data Models• Goal: Find structure in data• Applications– Finance– Engineering– Sciences

• Biological– Wherever we deal with data

• Some examples– Regression– Matrix factorization

• Challenges: Outliers and Missing data

Outliers Are Quite CommonGoogle search results for `male faces’

Need to Handle Outliers Properly

Noisy image Gaussian filtered image Desired result

Removing salt-and-pepper (outlier) noise

Missing Data Problem

Completing missing tracks

Incomplete tracksCompleted tracks by a sub-optimal method

Desired result

Missing tracks in structure from motion

Our Focus

• Outliers in regression– Linear regression– Kernel regression

• Matrix factorization in presence of missing data

Robust Linear Regression for High Dimension Problems

What is Regression?

• Regression– Find functional relation between y and x• x: independent variable• y: dependent variable

– Given • data: (yi,xi) pairs

• Model y = f(x, w)+n

– Estimate w– Predict y for a new x

Robust Regression

• Real world data corrupted with outliers• Outliers make estimates unreliable

• Robust regression– Unknown

• Parameter, w• Outliers

– Combinatorial problem• N data and k outliers

• C(N,k) ways

Prior Work

• Combinatorial algorithms– Random sample consensus (RANSAC)– Least Median Squares (LMedS)• Exponential in dimension

• M-estimators– Robust cost functions– local minima

Robust Linear Regression model

• Linear regression model : yi=xiTw+ei

– ei, Gaussian noise

• Proposed robust model: ei=ni+si

– ni, inlier noise (Gaussian)

– si, outlier noise (sparse)

• Matrix-vector form– y=Xw+n+s

• Estimate w, s

y1

y2

.

.yN

x1T

x2T

.

.xN

T

n1

n2

. .nN

s1

s2

. .sN

= + +

w1

w2

.wD

Simplification• Objective (RANSAC): Find w that minimizes the

number of outliers

• Eliminate w • Model: y=Xw+n+s • Premultiple by C : CX=0, N ≥ D– Cy=CXw+Cs+Cn– z=Cs+g

– g Gaussian• Problem becomes: • Solve for s -> identify outliers -> LS -> w

20 ||||||||min Cszss

tosubject

20 ||||||||min sXwysws,

tosubject

Relation to Sparse Learning

• Solve: – Combinatorial problem

• Sparse Basis Selection/ Sparse Learning• Two approaches :– Basis Pursuit (Chen, Donoho, Saunder 1995)– Bayesian Sparse Learning (Tipping 2001)

20 ||||||||min Cszss

tosubject

Basis Pursuit Robust regression (BPRR)

• Solve – Basis Pursuit Denoising (Chen et. al. 1995)– Convex problem– Cubic complexity : O(N3)

• From Compressive Sensing theory (Candes 2005)– Equivalent to original problem if

• s is sparse• C satisfy Restricted Isometry Property (RIP)

• Isometry: ||s1 - s2|| = ||C(s1 – s2)||• Restricted: to the class of sparse vectors

• In general, no guarantees for our problem

Cszss

thatsuch1

min

Bayesian Sparse Robust Regression (BSRR)

• Sparse Bayesian learning technique (Tipping 2001)– Puts a sparsity promoting prior on s : – Likelihood : p(z/s)=Ν(Cs,εI)– Solves the MAP problem p(s/z) – Cubic Complexity : O(N3)

N

i issp

1

1)(

Setup for Empirical Studies• Synthetically generated data

• Performance criteria– Angle between ground truth and estimated hyper-planes

Vary Outlier Fraction

BSRR performs well in all dimensions Combinatorial algorithms like RANSAC, MSAC, LMedS not

practical in high dimensions

Dimension = 2 Dimension = 8 Dimension = 32

Facial Age Estimation • Fgnet dataset : 1002 images of 82 subjects

• Regression– y : Age

– x: Geometric feature vector

Outlier Removal by BSRR

• Label data as inliers and outliers• Detected 177 outliers in 1002 images

BSRR

Inlier MAE 3.73

Outlier MAE 19.14

Overall MAE 6.45

•Leave-one-out testing

Summary for Robust Linear Regression

• Modeled outliers as sparse variable • Formulated robust regression as Sparse

Learning problem– BPRR and BSRR

• BSRR gives the best performance• Limitation: linear regression model– Kernel model

Robust RVM Using Sparse Outlier Model

Relevance Vector Machine (RVM)

• RVM model:– : kernel function

• Examples of kernels– k(xi, xj) = (xi

Txj)2 : polynomial kernel

– k(xi, xj) = exp( -||xi - xj||2/2σ2) : Gaussian kernel

• Kernel trick: k(xi,xj) = ψ(xi)Tψ(xj)– Map xi to feature space ψ(xi)

N

ii ewkwy

10),()( ixxx

),( ixxk

RVM: A Bayesian Approach• Bayesian approach– Prior distribution : p(w)– Likelihood :

• Prior specification– p(w) : sparsity promoting prior p(wi) = 1/|wi|– Why sparse?

• Use a smaller subset of training data for prediction• Support vector machine

• Likelihood – Gaussian noise

• Non-robust : susceptible to outliers

),|( wxyp

Robust RVM model

• Original RVM model– e, Gaussian noise

• Explicitly model outliers, ei= ni + si

– ni, inlier noise (Gaussian)

– si, outlier noise (sparse and heavy-tailed)

• Matrix vector form– y = Kw + n + s

• Parameters to be estimated: w and s

N

ijj ewkwy

10)( xx,

Robust RVM Algorithms

• y = [K|I]ws + n– ws = [wT sT]T : sparse vector

• Two approaches– Bayesian– Optimization

Robust Bayesian RVM (RB-RVM)

• Prior specification– w and s independent : p(w, s) = p(w)p(s)– Sparsity promoting prior for s: p(si)= 1/|si|

• Solve for posterior p(w, s|y) • Prediction: use w inferred above • Computation: a bigger RVM– ws instead of w

– [K|I] instead of K

Basis Pursuit RVM (BP-RVM)

• Optimization approach

– Combinatorial

• Closest convex approximation

• From compressive sensing theory– Same solution if [K|I] satisfies RIP• In general, can not guarantee

20 ||]|[||||||min ssw

wIKyws

tosubject

21 ||]|[||||||min ssw

wIKyws

tosubject

Experimental Setup

Prediction : Asymmetric Outliers Case

Image Denoising

• Salt and pepper noise– Outliers

• Regression formulation– Image as a surface over 2D grid• y: Intensity • x: 2D grid • Denoised image obtained by prediction

Salt and Pepper Noise

Some More ResultsRVM RB-RVM Median Filter

Age Estimation from Facial Images

• RB-RVM detected 90 outliers

• Leave-one-person-out testing

Summary for Robust RVM

• Modeled outliers as sparse variables• Jointly estimated parameter and outliers• Bayesian approach gives very good result

Limitations of Regression

• Regression: y = f(x,w)+n– Noise in only “y”– Not always reasonable

• All variables have noise– M = [x1 x2 … xN]– Principal component analysis (PCA)

• [x1 x2 … xN] = ABT – A: principal components– B: coefficients

– M = ABT: matrix factorization (our next topic)

Matrix Factorization in the presence of Missing Data

Applications in Computer Vision

• Matrix factorization: M=ABT

• Applications: build 3-D models from images– Geometric approach (Multiple views)

– Photometric approach (Multiple Lightings)

37

Structure from Motion (SfM)

Photometric stereo

Matrix Factorization

• Applications in Vision– Affine Structure from Motion (SfM)– Photometric stereo

• Solution: SVD– M=USVT

– Truncate S to rank r• A=US0.5, B=VS0.5

38

M =xij

yij= CST

Rank 4 matrix

M = NST, rank = 3

Missing Data Scenario

• Missed feature tracks in SfM

• Specularities and shadow in photometric stereo

39

Incomplete feature tracks

Challenges in Missing Data Scenario

• Can’t use SVD• Solve:

• W: binary weight matrix, λ: regularization parameter• Challenges

– Non-convex problem– Newton’s method based algorithm (Buchanan et. al. 2005)

• Very slow• Design algorithm

– Fast (handle large scale data) – Flexible enough to handle additional constraints

• Ortho-normality constraints in ortho-graphic SfM

)||||||(||||)(||min 222FFF

T BAABMWBA,

Proposed Solution

• Formulate matrix factorization as a low-rank semidefinite program (LRSDP)– LRSDP: fast implementation of SDP (Burer, 2001) • Quasi-Newton algorithm

• Advantages of the proposed formulation:– Solve large-scale matrix factorization problem– Handle additional constraints

41

Low-rank Semidefinite Programming (LRSDP)

• Stated as:• Variable: R• Constants• C: cost • Al, bl: constants

• Challenge• Formulating matrix factorization as LRSDP• Designing C, Al, bl

klbtosubject lT

lT ,...,2,1,min RRARRC

R

Matrix factorization as LRSDP: Noiseless Case

• We want to formulate:

• As:• LRSDP formulation:

),()(||||||||min ,,22

,jifortosubject jiji

TFF

BAMABBA

)(||||||||

)(||||),(||||22

22

TFF

TF

TF

trace

tracetrace

RRBA

BBBAAA

jimjiT

jijiT

,,,, )()( MRRMAB

||,...,2,1, lbtosubject lTT RRARRC l

C identity matrix,Al indicator matrix

Affine SfM

• Dinosaur sequence

• MF-LRSDP gives the best reconstruction

72% missing data

Photometric Stereo

• Face sequence

• MF-LRSDP and damped Newton gives the best result

42% missing data

Additional Constraints:Orthographic Factorization

• Dinosaur sequence

Summary

• Formulated missing data matrix factorization as LRSDP– Large scale problems– Handle additional constraints

• Overall summary– Two statistical data models• Regression in presence of outliers

– Role of sparsity• Matrix factorization in presence of missing data

– Low rank semidefinite program

Thank you! Questions?

48