Tackling Box-Constrained Optimization Via a New Projected Quasi-Newton Approach

Dongmin Kim, Suvrit Sra, Inderjit Dhillon

Abstract:   Numerous scienti c applications across a variety of elds depend on box-constrained convex optimization. Box-constrained problems therefore continue to attract research interest. We address box-constrained (strictly convex) problems by deriving two new quasi-Newton algorithms. Our algorithms are positioned between the projected-gradient [J. B. Rosen, J. SIAM, 8(1), 1960, pp. 181{217], and projected-Newton [D. P. Bertsekas, SIAM J. Cont. & Opt., 20(2), 1982, pp. 221{246] methods. We also prove their convergence under a simple Armijo step-size rule. We provide experimental results for two particular box-constrained problems: nonnegative least squares (NNLS), and nonnegative Kullback-Leibler (NNKL) minimization. For both NNLS and NNKL our algorithms perform competitively against well-established methods on medium-sized problems; for larger problems our approach frequently outperforms the competition.

Download: pdf

Citation

  • Tackling Box-Constrained Optimization Via a New Projected Quasi-Newton Approach (pdf, software)
    D. Kim, S. Sra, I. Dhillon.
    SIAM Journal on Scientific Computing 30(6), pp. 3548-3563, December 2010.

    Bibtex: