12.07.2015 Views

v2010.10.26 - Convex Optimization

v2010.10.26 - Convex Optimization

v2010.10.26 - Convex Optimization

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

298 CHAPTER 4. SEMIDEFINITE PROGRAMMING⎛⎡y ⋆ ⎜⎢= round⎝⎣1 − 1 √21 − 1 √21⎤⎞⎥⎟⎦⎠ =⎡⎢⎣001⎤⎥⎦ (700)(infeasible, with or without rounding, with respect to original problem (680))whereas solving semidefinite program (692) produces⎡⎤1 1 −1 1round(G ⋆ ) = ⎢ 1 1 −1 1⎥⎣ −1 −1 1 −1 ⎦ (701)1 1 −1 1with sorted eigenvaluesλ(G ⋆ ) =⎡⎢⎣3.999999650572640.00000035942736−0.00000000000000−0.00000001000000⎤⎥⎦ (702)Truncating all but the largest eigenvalue, from (688) we obtain (confer y ⋆ )⎛⎡⎤⎞⎡ ⎤0.99999999625299 1x ⋆ = round⎝⎣0.99999999625299 ⎦⎠ = ⎣ 1 ⎦ (703)0.00000001434518 0the desired minimal cardinality Boolean result.4.2.3.1.3 Exercise. Minimal cardinality Boolean art.Assess general performance of standard-practice approximation (697) ascompared with the proposed semidefinite program (692). 4.2.3.1.4 Exercise. Conic independence.Matrix A from (682) is full-rank having three-dimensional nullspace. Findits four conically independent columns. (2.10) To what part of proper coneK = {Ax | x ≽ 0} does vector b belong?4.2.3.1.5 Exercise. Linear independence.Show why fat matrix A , from compressed sensing problem (518) or (523),may be regarded full-rank without loss of generality. In other words: Is aminimal cardinality solution invariant to linear dependence of rows?

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!