Lossy compression for PDE-constrained optimization: adaptive error control
For the solution of optimal control problems governed by nonlinear parabolic PDEs, methods working on the reduced objective functional are often employed to avoid solving large systems in the dimension of the full spatio-temporal discretization. The evaluation of the reduced gradient requires one solve of the state equation forward in time, and one backward solve of the adjoint equation. The state enters into the adjoint equation, requiring the storage of a full 4D data set. If Newton-CG methods are used, two additional trajectories have to be stored. To get numerical results which are accurate enough, in many cases very fine discretizations in time and space are necessary, which leads to a significant amount of data to be stored and transmitted to mass storage. Lossy compression methods were developed to overcome the storage problem by reducing the accuracy of the stored trajectories. The inexact data induces errors in the reduced gradient and reduced Hessian. In this paper, we analyze the influence of such a lossy trajectory compression method on Newton-CG methods for optimal control of parabolic PDEs and design an adaptive strategy for choosing appropriate quantization tolerances.
KeywordsOptimal control Semi-linear parabolic PDEs Newton-CG Trajectory storage Lossy compression
Mathematics Subject Classification35K58 49M15 65M60 68P30 94A29
The authors gratefully acknowledge support by the DFG Research Center Matheon, project F9.
- 2.Britton, N.F.: Reaction-Diffusion Equations and Their Application to Biology. Academic Press, Waltham (1986)Google Scholar
- 3.Deuflhard, P., Weiser, M.: Local inexact Newton multilevel FEM for nonlinear elliptic problems. In: Bristeau, M.O., Etgen, G., Fitzgibbon, W., Lions, J.L., Periaux, J., Wheeler, M. (eds.) Computational Science for the 21st Century, pp. 129–138. Wiley, Hoboken (1997)Google Scholar
- 4.Du, X., Haber, E., Karampataki, M., Szyld, D.B.: Varying iteration accuracy using inexact conjugate gradients in control problems governed by PDEs. In: R. Chen (ed.) Proceedings of the 2nd Annual Conference on Computational Mathematics, Computational Geometry & Statistics (CMCGS 2013), pp. 29–38 (2013)Google Scholar
- 6.Götschel, S., Chamakuri, N., Kunisch, K., Weiser, M.: Lossy compression in optimal control of cardiac defibrillation. J. Sci. Comput. 60(1), 35–59 (2014)Google Scholar
- 7.Götschel, S., von Tycowicz, C., Polthier, K., Weiser, M.: Reducing memory requirements in scientific computing and optimal control. In: Carraro, T., Geiger, M., Körkel, S., Rannacher, R. (eds.) Multiple Shooting and Time Domain Decomposition Methods. Springer (2014, to appear)Google Scholar
- 14.Kunisch, K., Wagner, M.: Optimal control of the bidomain system (I): the monodomain approximation with the Rogers–McCulloch model. Nonlinear Anal. Real World Appl. 13(4), 1525–1550 (2012).Google Scholar
- 15.Meurant, G.A.: The Lanczos and Conjugate Gadient Algorithms: From Theory to Finite Precision Computations, vol. 19. SIAM, Philadelphia (2006)Google Scholar
- 17.Nagaiah, C., Kunisch, K., Plank, G.: Numerical solution for optimal control of the reaction-diffusion equations in cardiac electrophysiology. Comput. Optim. Appl. 49, 149–178 (2011). doi: 10.1007/s10589-009-9280-3.
- 22.Sleijpen, G.L.G., van der Vorst, H.A.: Reliable updated residuals in hybrid Bi-CG methods. Computing 56(2), 141–163 (1996)Google Scholar
- 26.van der Vorst, H.A., Ye, Q.: Residual replacement strategies for Krylov subspace iterative methods for the convergence of true residuals. SIAM J. Sci. Comput. 22(3), 835–852 (2000)Google Scholar