COM Optimization for Communications 8. Semidefinite Programming

Similar documents
ELE539A: Optimization of Communication Systems Lecture 15: Semidefinite Programming, Detection and Estimation Applications

Semidefinite Programming Basics and Applications

Convex relaxation. In example below, we have N = 6, and the cut we are considering

Convex relaxation. In example below, we have N = 6, and the cut we are considering

Multiuser Downlink Beamforming: Rank-Constrained SDP

Lecture: Examples of LP, SOCP and SDP

Relaxations and Randomized Methods for Nonconvex QCQPs

ELEC E7210: Communication Theory. Lecture 10: MIMO systems

Convex Optimization. (EE227A: UC Berkeley) Lecture 6. Suvrit Sra. (Conic optimization) 07 Feb, 2013

arxiv: v1 [math.oc] 23 Nov 2012

Acyclic Semidefinite Approximations of Quadratically Constrained Quadratic Programs

IE 521 Convex Optimization

Convex Optimization M2

Preliminaries Overview OPF and Extensions. Convex Optimization. Lecture 8 - Applications in Smart Grids. Instructor: Yuanzhang Xiao

Introduction to Semidefinite Programming I: Basic properties a

CSCI 1951-G Optimization Methods in Finance Part 10: Conic Optimization

There are several approaches to solve UBQP, we will now briefly discuss some of them:

Introduction to Convex Optimization

Optimization in Wireless Communication

A RLT relaxation via Semidefinite cut for the detection of QPSK signaling in MIMO channels

Nonconvex Quadratic Optimization, Semidefinite Relaxation, and Applications

Agenda. Applications of semidefinite programming. 1 Control and system theory. 2 Combinatorial and nonconvex optimization

Lecture 4: January 26

Convex Optimization and Its Applications in Signal Processing

Per-Antenna Power Constrained MIMO Transceivers Optimized for BER

Solution to EE 617 Mid-Term Exam, Fall November 2, 2017

SPECTRUM SHARING IN WIRELESS NETWORKS: A QOS-AWARE SECONDARY MULTICAST APPROACH WITH WORST USER PERFORMANCE OPTIMIZATION

New Rank-One Matrix Decomposition Techniques and Applications to Signal Processing

Handout 6: Some Applications of Conic Linear Programming

Lecture 4: Linear and quadratic problems

MIT Algebraic techniques and semidefinite optimization February 14, Lecture 3

Semidefinite Relaxation and Its Applications in Signal Processing and Communications

ELE539A: Optimization of Communication Systems Lecture 6: Quadratic Programming, Geometric Programming, and Applications

Semidefinite Relaxation of Nonconvex Quadratic Optimization: A Key Technique in Signal Processing and Communications

Applications of Robust Optimization in Signal Processing: Beamforming and Power Control Fall 2012

Semi-Definite Programming (SDP) Relaxation Based Semi-Blind Channel Estimation for Frequency-Selective MIMO MC-CDMA Systems

Lecture 7 MIMO Communica2ons

Advances in Convex Optimization: Theory, Algorithms, and Applications

Limited Feedback in Wireless Communication Systems

Convex Optimization and Modeling

USING multiple antennas has been shown to increase the

Acyclic Semidefinite Approximations of Quadratically Constrained Quadratic Programs

QoS Constrained Robust MIMO Transceiver Design Under Unknown Interference

Simultaneous SDR Optimality via a Joint Matrix Decomp.

Vector Channel Capacity with Quantized Feedback

Capacity Region of the Two-Way Multi-Antenna Relay Channel with Analog Tx-Rx Beamforming

Comparing Convex Relaxations for Quadratically Constrained Quadratic Programming

Sum-Power Iterative Watefilling Algorithm

Interior Point Methods: Second-Order Cone Programming and Semidefinite Programming

A Continuation Approach Using NCP Function for Solving Max-Cut Problem

Lecture: Convex Optimization Problems

Space-Time Coding for Multi-Antenna Systems

Single-User MIMO systems: Introduction, capacity results, and MIMO beamforming

HEURISTIC METHODS FOR DESIGNING UNIMODULAR CODE SEQUENCES WITH PERFORMANCE GUARANTEES

Semidefinite Relaxation of Quadratic Optimization Problems

Exercises. Exercises. Basic terminology and optimality conditions. 4.2 Consider the optimization problem

Handout 6: Some Applications of Conic Linear Programming

Exploiting Partial Channel Knowledge at the Transmitter in MISO and MIMO Wireless

BLOCK DATA TRANSMISSION: A COMPARISON OF PERFORMANCE FOR THE MBER PRECODER DESIGNS. Qian Meng, Jian-Kang Zhang and Kon Max Wong

On Improving the BER Performance of Rate-Adaptive Block Transceivers, with Applications to DMT

Cognitive MIMO Radio: Incorporating Dynamic Spectrum Access in Multiuser MIMO Network

Achievable Outage Rate Regions for the MISO Interference Channel

Global Quadratic Minimization over Bivalent Constraints: Necessary and Sufficient Global Optimality Condition

Convex Optimization in Communications and Signal Processing

An Introduction to Linear Matrix Inequalities. Raktim Bhattacharya Aerospace Engineering, Texas A&M University

3. ESTIMATION OF SIGNALS USING A LEAST SQUARES TECHNIQUE

Example 1 linear elastic structure; forces f 1 ;:::;f 100 induce deections d 1 ;:::;d f i F max i, several hundred other constraints: max load p

Problem structure in semidefinite programs arising in control and signal processing

Capacity optimization for Rician correlated MIMO wireless channels

Novel spectrum sensing schemes for Cognitive Radio Networks

Tutorial on Convex Optimization: Part II

DOWNLINK transmit beamforming has recently received

Incremental Coding over MIMO Channels

7.1 Linear Systems Stability Consider the Continuous-Time (CT) Linear Time-Invariant (LTI) system

Optimal Transmit Strategies in MIMO Ricean Channels with MMSE Receiver

On the Sandwich Theorem and a approximation algorithm for MAX CUT

EE Applications of Convex Optimization in Signal Processing and Communications Dr. Andre Tkacenko, JPL Third Term

4. Convex optimization problems

Maximum Achievable Diversity for MIMO-OFDM Systems with Arbitrary. Spatial Correlation

2318 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 6, JUNE Mai Vu, Student Member, IEEE, and Arogyaswami Paulraj, Fellow, IEEE

Optimization of Multistatic Cloud Radar with Multiple-Access Wireless Backhaul

The maximal stable set problem : Copositive programming and Semidefinite Relaxations

Convex Optimization for Signal Processing and Communications: From Fundamentals to Applications

SINR Balancing in the Downlink of Cognitive Radio Networks with Imperfect Channel Knowledge

EE5138R: Problem Set 5 Assigned: 16/02/15 Due: 06/03/15

CSC Linear Programming and Combinatorial Optimization Lecture 10: Semidefinite Programming

Lagrange Duality. Daniel P. Palomar. Hong Kong University of Science and Technology (HKUST)

CHARACTERIZATION of the capacity region of an Interference. On the Sum-Rate of MIMO Interference Channel

LETTER A Semidefinite Relaxation Approach to Spreading Sequence Estimation for DS-SS Signals

Canonical Problem Forms. Ryan Tibshirani Convex Optimization

Sum Rate Maximizing Multigroup Multicast Beamforming under Per-antenna Power Constraints

Multi-antenna Relay Network Beamforming Design for Multiuser Peer-to-Peer Communications

Lecture 20: November 1st

Minimum Mean Squared Error Interference Alignment

Schur-convexity of the Symbol Error Rate in Correlated MIMO Systems with Precoding and Space-time Coding

THE USE OF optimization methods is ubiquitous in communications

A SEMIDEFINITE RELAXATION BASED CONSERVATIVE APPROACH TO ROBUST TRANSMIT BEAMFORMING WITH PROBABILISTIC SINR CONSTRAINTS

Agenda. 1 Cone programming. 2 Convex cones. 3 Generalized inequalities. 4 Linear programming (LP) 5 Second-order cone programming (SOCP)

Lecture 5: Antenna Diversity and MIMO Capacity Theoretical Foundations of Wireless Communications 1. Overview. CommTh/EES/KTH

6.854J / J Advanced Algorithms Fall 2008

Transcription:

COM524500 Optimization for Communications 8. Semidefinite Programming Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 1 Semidefinite Programming () Inequality form: min c T x s.t. F(x) 0 where F(x) = F 0 + x 1 F 1 +... + x n F n, F i S p p. Standard form: min tr(cx) s.t. X 0 tr(a i X) = b i, i = 1,..., m where A i S n n, and C S n n. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 2

The inequality & standard forms can be shown to be equiv. F(x) 0 is commonly known as a linear matrix inequality (LMI). An with multiple LMIs min c T x s.t. F i (x) 0, i = 1,..., m can be reduced to an with one LMI since F i (x) 0, i = 1,..., m blkdiag(f 1 (x),..., F m (x)) 0 where blkdiag is the block diagonal operator. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 3 Example: Max. eigenvalue minimization Let λ max (X) denote the maximum eigenvalue of a matrix X. Max. eigenvalue minimization problem: min x λ max (A(x)) where A(x) = A 0 + x 1 A 1 +... + x n A n. We note that fixing x, λ max (A(x)) t A(x) ti 0 Hence, the problem is equiv. to min x,t t s.t. A(x) ti 0 Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 4

Standard LP: LP as min c T x s.t. x 0, a T i x = b i, i = 1,..., m Let C = diag(c), & A i = diag(a i ). The standard min tr(cx) s.t. X 0 tr(a i X) = b i, i = 1,..., m is equiv. to the LP since X 0 = diag(x) 0. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 5 Inequality form LP min c T x s.t. Ax b is equiv. to the min c T x s.t. diag(ax b) 0 because X 0 = X ii 0 for all i. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 6

Schur Complements Let X S n and partition X = A B T B C S = C B T A 1 B is called the Schur complement of A in X (provided A 0). Important facts: X 0 iff A 0 and S 0. If A 0, then X 0 iff S 0. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 7 Schur complements are useful in turning some nonlinear constraints into LMIs: Example: The convex quadratic inequality (Ax + b) T (Ax + b) c T x d 0 is equivalent to I (Ax + b) T Ax + b 0 c T x + d Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 8

QCQP as A convex QCQP can always be written as min A 0 x + b 0 2 2 ct 0 x d 0 s.t. A i x + b i 2 2 c T i x d i 0, i = 1,..., L By Schur complement, the QCQP is equiv. to min t [ ] s.t. I A 0 x + b 0 0 (A 0 x + b 0 ) T [ c T 0 x + d 0 + t ] I A i x + b i 0, i = 1,..., L (A i x + b i ) T c T i x + d i Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 9 Example: The second order cone inequality: Ax + b 2 f T x + d If the domain is such that f T x + d > 0, the inequality can be re-expressed as f T x + d 1 f T x + d (Ax + b)t (Ax + b) 0. By Schur complement, the inequality is equiv. to (ft x + d)i Ax + b 0 (Ax + b) T f T x + d This result indicates that SOCP can be turned to an. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 10

Application in Combinatorial Optimization Problem Statement: Consider the Boolean quadratic program (BQP): max x T Cx s.t. x i { 1, +1}, where C S n. The BQP is (very) nonconvex since the equality constraints x 2 i = 1 are nonconvex. In fact the BQP is NP-hard in general. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 11 Practical Example I: MAXCUT [GW95] Input: A graph G = (V, E) with weights w ij for (i, j) E. Assume w ij 0 and w ij = 0 if (i, j) / E. Goal: Divide nodes into two parts so as to maximize the weight of the edges whose nodes are in different parts. 1 w 13 3 w 14 4 2 w 25 5 Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 12

Suppose V = {1, 2,..., n}. The MAXCUT problem takes the form n n 1 x i x j max w ij 2 i=1 j=i+1 s.t. x i { 1, +1}, which can be rewritten as max x T Cx s.t. x i { 1, +1}, where C ij = C ji = 1 4 w ij for i j, & C ii = 1 2 n j=1 w ij. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 13 Practical Example II: ML MIMO detection [MDW + 02] H 11 H 1,n data symbols Multiplexer H m,1 Demultiplexer H m,n The tx & rx sides have n & m antennas, respectively. We consider a spatial multiplexing system, where each tx antenna transmits its own sequence of symbols. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 14

Assume frequency flat fading, & antipodal modulation. The received signal model may be expressed as y = Hs + v where y R m s { 1, +1} n H R m n v R n multi-receiver output vector; transmitted symbols; MIMO (or multi-antenna) channel; Gaussian noise with zero mean & covariance σ 2 I. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 15 Maximum-likelihood (ML) detection of s: min s {±1} n y Hs 2 2 = min s {±1} n st H T Hs 2s T H T y The ML problem is a nonhomogeneous BQP. But it can be reformulated as a homogeneous BQP. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 16

The ML problem can be homogenized as follows: min s {±1} n st H T Hs 2s T H T y = min s {±1} n c {±1} (c s) T H T H(c s) 2(c s) T H T y (s = c s) = min s {±1} n s T H T H s 2c s T H T y (c 2 = 1) c {±1} [ ] = min s T c HT H H T y ( s,c) {±1} n+1 y T H 0 c Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 17 BQP Approximation by : The BQP can be reformulated as max x,x tr(cx) s.t. X = xx T X ii = 1, Now, the objective function is linear for any C; the {±1} constraints on x becomes linear in X; but X = xx T is nonconvex. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 18

Using the fact that X = xx T = X 0 we approximate the BQP by solving the following : max tr(cx) s.t. X 0 X ii = 1, Once the is solved, its solution is used to approximate the BQP solution (e.g., by applying rank-1 approximation to the solution). Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 19 approximation looks heuristic, but it has been shown to provide good approx. accuracy. To see this, let f BQP = max x {±1} n xt Cx, f = max X 0,X ii =1 i trcx Goemans & Williamson [GW95] showed that for MAXCUT where C ij 0 for i j & C 0, 0.87856f f BQP f Nesterov [N98] showed that for C 0, 2 π f f BQP f Zhang [Z00] showed that if C ij 0 for i j, f BQP = f Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 20

Tx Downlink Beamforming for Broadcasting The problem is the same as that in the last lecture, except that the transmitter sends common information to all receivers. The received signal remains the same: y i = h T i x + v i, but the transmitted signal is now given by x = fs where f C m is the tx beamformer vector & s C is the information bearing signal. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 21 Problem: [SDL06] Minimize the tx power, subject to constraints that the received SNR is no less than some pre-specified threshold: h T i f 2 2 σ 2 i γ o, where γ o is the SNR threshold, & σ 2 i is the noise variance. Let Q i = 1 γ o h σi 2 i ht i. The problem can be written as min f 2 2 s.t. tr(ff H Q i ) 1, Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 22

Unfortunately the tx beamformer design problem here is nonconvex. Even worse, the problem is shown to be NP-hard [SDL06]. However, we can approximate the problem using. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 23 An equivalent form of the problem: min tr(f) f C m,f H m s.t. F = ff H tr(fq i ) 1, approximation: min tr(f) s.t. F 0 tr(fq i ) 1, Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 24

Another Problem: Maximize the weakest received SNR, subject to a constraint that the tx power is no greater than a threshold P o : max min i=1,...,n 1 σ 2 i s.t. f 2 2 P o This problem is also NP-hard. h i f 2 Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 25 The problem can be re-expressed as max t s.t. 1 σ 2 i h T i f 2 t, f 2 2 P o The associated approximation: max t s.t. 1 tr(h σi 2 i ht i F 0 F) t, tr(f) P o Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 26

References [MDW + 02] [GW95] [N98] [Z00] [SDL06] W.-K. Ma, T.N. Davidson, K.M. Wong, P.C. Ching, & Z.-Q. Luo, Quasi-ML multiuser detection using SDR with application to sync. CDMA, IEEE Trans. Signal Proc., 2002. M.X. Goemans & D.P. Williamson, Improved approx. alg. for max. cut & satisfiability problem using, J. ACM, 1995. Y.E. Nesterov, SDR and nonconvex quadratic opt., Opt. Meth. Software, 1998. S. Zhang, Quadratic max. and SDR, Math. Program., 2000. N.D. Sidiropoulos, T.N. Davidson, & Z.-Q. Luo, Transmit beamforming for physical layer multicasting, IEEE Trans. Signal Proc., 2006. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 27 Additional Reading on Applications [DLW00] [DTS01] T.N. Davidson, Z.-Q. Luo, & K.M. Wong, Design of orthogonal pulse shapes for communications via, IEEE Trans. Signal Proc., 2000. B. Dumitrescu, I. Tabus, & P. Stoica, On the parameterization of positive real sequences & MA parameter estimation, IEEE Trans. Signal Proc., 2001. Institute Comm. Eng. & Dept. Elect. Eng., National Tsing Hua University 28