Scribd
Upload
69 views

Tensor Decomp Presentation

This document summarizes work on matrix and tensor completion. It discusses matrix completion methods like convex relaxation using nuclear norm minimization and alternating least squares. For tensors, it outlines tensor decomposition challenges and existing approaches like alternating least squares and convex relaxations using new tensor norms. The author proposes further empirical study of recent tensor norm methods and attempts to prove theoretical bounds, construct better convex relaxations, and generalize results about local optima from matrices to tensors.

Uploaded by

Mahbod Matt Olfat
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
69 views

Tensor Decomp Presentation

This document summarizes work on matrix and tensor completion. It discusses matrix completion methods like convex relaxation using nuclear norm minimization and alternating least squares. For tensors, it outlines tensor decomposition challenges and existing approaches like alternating least squares and convex relaxations using new tensor norms. The author proposes further empirical study of recent tensor norm methods and attempts to prove theoretical bounds, construct better convex relaxations, and generalize results about local optima from matrices to tensors.

Uploaded by

Mahbod Matt Olfat
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Matrix Completion and Tensor

Decomposition and Completion


Matt Olfat
UC Berkeley IEOR

Overview
Matrix(comple.on(mo.va.on(and(methods(
Background(on(tensors(and(basic(generaliza.ons(from(matrices(
Overview(of(challenges(and(methods(for(tensor(decomposi.on(
Exis.ng(tensor(comple.on(methods(
Proposed(works(
10/5/16

Olfat, M.

Matrix Sensing & Completion


Problem:

Minimize rank of matrix X subject to affine constraints A(X) = b


In the case that A is a sampling operator, called Matrix Completion
More general case called Matrix Sensing, contains vector cardinality minimization

Motivation:

Motivated by compressed sensing, Cands & Tao (2005)


Popularized by the Netflix Prize, recommending movies based on sparse ratings
Applications include image compression, sensor positioning, multiclass learning

Challenges:
a) X in the null space of A:
2
3
1 0 0
e1 eT1 = 40 0 05
0 0 0
10/5/16

b) Matrix rank highly non-convex:

1 0
0 0

+ (1 )
=
0 0
0 1
0 1
Olfat, M.

Matrix Sensing & Completion Approaches


Convex Relaxation: rank(X) ! kXk

Recht, Fazel & Parrilo (2010) show nuclear norm ball is tightest convex relaxation of rank ball
Define the r-isometry constant: r = min : (1
)kXkF kA(X)k (1 + )kXkF 8X : rank(X) r
Show min kXk gives exact solution given Restricted Isometry Property: 5r < 1/10
Give cases where RIP holds with high probability
Candes & Recht (2009) build on nuclear norm relaxation specifically for matrix completion
Can give exact solution given O(rn1.2 log n) samples for small r, or O(rn1.25 log n) samples for any r
Constant dependent on coherence of the matrix: (X) = nr max1in kPX ei k2

Alternating Least Squares: X := U V T ! min kX

Ut VtT k2F

Iterate through Ut+1 = minU kX U VtT k2 , Vt+1 = minV kX Ut+1 V T k2


In the case of matrix completion, objective becomes min kb A(Ut VtT )k2F
Initialize via singular vectors of clipped SVD
Uses less memory and converges faster than the convex relaxation method
Jain, Netrapalli & Sanghavi (2013) prove first theoretical bounds on ALS
F
For matrix sensing, t > 2 log kXk
! kX Ut VtT kF

rkXkF
kXk
For matrix completion, t = O(log F ) ! kX Ut VtT kF if |A| = O(4 (X)r4.5 n log n log )
Hardt (2013) improves results by a factor of r4 (X)5 for matrix completion

10/5/16(

Olfat,(M.(

4(

Tensor Decomposition
Origination:

Applications:

First proposed by Hitchcock (1927a, b)


First multi-way model using tensors by Cattel
(1944a, b) in the field of psychometrics
Employed sporadically throughout 1960s
Spread to chemistry, Appellof & Davidson (1981)
Of more general interest in 2000s

Signal processing
Numerical linear algebra
Image compression
Data mining
Neuroscience
Quantum physics
Full survey by Kolda & Bader (2009)

Generalizations from Matrices:

A tensor is a multi-dimensional array X 2 RI1 I2 IN


P
P
Space equipped with inner product hX , Yi = i1 2[I1 ] iN 2[IN ] xi1 ...iN yi1 ...iN
Rank-one tensor defined as X = a(1) a(N ) , a(i) 2 RIi 8i
P
(N )
Standard is CANDECOMP-PARAFAC (CP): X [ , A(1) , . . . , A(N ) ] = i2[r] r a(a)
ar
r
Generalization of SVD in matrices
Unlike matrix decompositions, CP decomposition is often unique (specifics in Kruskal (1989))
Rank of a tensor is the rank of its lowest-rank CP decomposition
Higher order SVD and Power Methods proposed by De Latheauwer et al. (2000a, b)

10/5/16

Olfat, M.

Tensor Decomposition
Some Problems:

Hillar & Lim (2012) show that finding most standard decompositions for tensors is NP-hard
Clipping smallest components in CP decomposition gives low-rank approximation, but not always the best
Finding rank is therefore NP-hard in general as well
In fact, De Silva & Lim (2008) show that finding low-rank approximations of tensors is ill-posed
However, also show use of hyper-determinant in finding tensor rank (more later)
Decomposition can depend on whether tensor can take complex numbers:

1 0
0 1
X 2 R222 ! rank(X ) = 3; X 2 C222 ! rank(X ) = 2
X1 =
,X2 =
0 1
1 0

Approaches:

Anandkumar et al. (2012) efficiently solve for low CP-rank orthogonal symmetric tensor via power method
Anandkumar et al. (2014) efficiently find approximations for low CP-rank incoherent tensors
Generally depend on whitening step to make tensor symmetric first, but this is computationally expensive
Ge et al. (2015) suggest online stochastic gradient descent method for decomposition
But only provably converges to local solution
In the case of Matrix Completion, Ge et al. (2016) showed that all local optima are global optima when
restricting set of matrices to be PSD
Raises similar questions for tensors, but first need to rigorously define PSD tensors..

10/5/16

Olfat, M.

Tensor Completion Approaches


Alternating Least Squares:
minX
s.t.

rank(X )
A(X ) = b

! min kA(X )

A(

i2[r]

(1)
i (ai

(N )

ai

))k2F

Generalization of ALS for matrix completion proposed by Jain & Oh (2014)


Initialize via iterations of tensor power method
Generally cannot guarantee local optimality, but can show initialization starts off close to global optimum
Show |A| = O(6 4 (X)r5 n 1/5 (log n)4 log rkX kF ) ! error for a third-order tensor
However, only works for low-rank orthogonal decompositions, has sub-optimal dependence on tensor rank

Convex Relaxation: rank(X ) ! kX k ?

New line of work seeks to design further relaxations of matrix rank ball via new norms or decompositions
Rauhut & Stojanac (2015) construct the k -norm based on concepts from computational algebraic geometry
Specifically, design nested sets based on relaxations of polynomial ideal generated by hyper-determinant of
tensor, which provably converge to convex hull of original ideal
Use Grbner basis to formulate SDP problem to efficiently minimize norms under affine constraints
Use norm of order 1 to recover third-order tensor, but do not provide theoretical bounds (next slide)
Nie & Wang (2014) use similar approach to recover best rank-1 approximations
Aswani (2016) also defines new decomposition, objective function that allows randomized approach

10/5/16

Olfat, M.

Results from Theta Norm Approach


Promising results, merit further investigation

10/5/16

Olfat, M.

Proposed Works
Conduct more extensive empirical
study of recently proposed norm
relaxation methods for tensors

Attempt to prove theoretical


bounds
Based on previous task

Bounds do not currently exists


But methods provide promising results

Attempt to construct better


convex relaxations
One option is to build on those currently
proposed
Another is to try to generalize methods
from integer programming
Otherwise, attempt to show that exact
generalizations cannot exist
10/5/16

Attempt to generalize results


regarding local optima of matrices
to tensors
To help validate or simplify existing ALS
methods
Cut out costly initialization step

Olfat, M.

You might also like