Abstract
We study the method of cyclic projections when applied to closed and linear subspaces \({M_{i}}, i=1,\ldots, m\), of a real Hilbert space \(\mathcal {H}\). We show that the average distance to individual sets enjoys a polynomial behavior \(o(k^{-1/2})\) along the trajectory of the generated iterates. Surprisingly, when the starting points are chosen from the subspace \(\sum _{i=1}^{m}M_i^{\perp }\), our result yields a polynomial rate of convergence \(\mathcal {O}(k^{-1/2})\) for the method of cyclic projections itself. Moreover, if \(\sum _{i=1}^{m} M_i^{\perp }\) is not closed, then both of the aforementioned rates are best possible in the sense that the corresponding polynomial \(k^{1/2}\) cannot be replaced by \(k^{1/2+\varepsilon }\) for any \({\varepsilon } >0\).
Similar content being viewed by others
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
Aragón Artacho, F.J., Campoy, R.: Optimal rates of linear convergence of the averaged alternating modified reflections method for two subspaces. Numer. Algorithms. 82, 397–421 (2019)
Aronszajn, N.: Theory of reproducing kernels. Trans. Amer. Math. Soc. 68, 337–404 (1950)
Badea, C., Grivaux, S., Müller, V.: The rate of convergence in the method of alternating projections. Algebra i Analiz. 23, 1–30 (2011)
Badea, C., Seifert, D.: Ritt operators and convergence in the method of alternating projections. J. Approx. Theory. 205, 133–148 (2016)
Badea, C., Seifert, D.: Quantified asymptotic behaviour of Banach space operators and applications to iterative projection methods. Pure Appl. Funct. Anal. 2, 585–598 (2017)
Bargetz, C., Reich, S., Zalas, R.: Convergence properties of dynamic string-averaging projection methods in the presence of perturbations. Numer. Algorithms. 77, 185–209 (2018)
Bauschke, H.H., Cruz, J.Y.B., Nghia, T.T.A., Phan, H.M., Wang, X.: The rate of linear convergence of the Douglas-Rachford algorithm for subspaces is the cosine of the Friedrichs angle. J. Approx. Theory. 185, 63–79 (2014)
Bauschke, H.H., Cruz, J.Y.B., Nghia, T.T.A., Phan, H.M., Wang, X.: Optimal rates of linear convergence of relaxed alternating projections and generalized Douglas-Rachford methods for two subspaces. Numer. Algorithms. 73, 33–76 (2016)
Bauschke, H.H., Borwein, J.M.: On projection algorithms for solving convex feasibility problems. SIAM Rev. 38, 367–426 (1996)
Bauschke, H.H., Borwein, J.M., Lewis, A.S.: The method of cyclic projections for closed convex sets in Hilbert space, in Recent developments in optimization theory and nonlinear analysis (Jerusalem 1995). Contemp. Math. Amer. Math. Soc. Providence. RI. 204, 1–38 (1997)
Bauschke, H.H., Deutsch, F., Hundal, H.: Characterizing arbitrarily slow convergence in the method of alternating projections. Int. Trans. Oper. Res. 16, 413–425 (2009)
Borodin, P.A., Kopecká, E.: Alternating projections, remotest projections, and greedy approximation. J. Approx. Theory. 260, 105486 (2020)
Brezis, H.: Functional Analysis, Sobolev Spaces and Partial Differential Equations. Universitext. Springer, New York (2011)
Cegielski, A.: Iterative methods for fixed point problems in Hilbert spaces, Lecture Notes in Mathematics, vol. 2057. Springer, Heidelberg (2012)
Cohen, G.: Iterates of a product of conditional expectation operators. J. Funct. Anal. 242, 658–668 (2007)
Crouzeix, M.: A functional calculus based on the numerical range: applications. Linear Multilinear Algebra. 56, 81–103 (2008)
Deutsch, F.: Best approximation in inner product spaces, vol. 7. Springer Verlag, New York (2001)
Deutsch, F., Hundal, H.: Slow convergence of sequences of linear operators II: arbitrarily slow convergence. J. Approx. Theory. 162, 1717–1738 (2010)
Deutsch, F., Hundal, H.: Arbitrarily slow convergence of sequences of linear operators: a survey, in Fixed-point algorithms for inverse problems in science and engineering, vol. 49, pp. 213–242. Appl. Springer, New York (2011)
Deutsch, F,. Hundal, H.: Arbitrarily slow convergence of sequences of linear operators, in Infinite products of operators and their applications, vol. 636, pp. 93–120. Contemp. Math. Amer. Math. Soc. Providence. RI (2015)
Evron, I., Moroshko, E., Ward, R., Srebro, N., Soudry, D.: How catastrophic can catastrophic forgetting be in linear regression? Proceedings of the thirty-fifth conference on learning theory. 178, 728–4079 (2022)
Franchetti, C., Light, W.: On the von neumann alternating algorithm in Hilbert space. J. Math. Anal. Appl. 114, 305–314 (1986)
Galántai, A.: Projectors and projection methods, vol. 6. Kluwer Academic Publishers. Boston, MA (2004)
Halperin, I.: The product of projection operators. Acta Sci. Math. Szeged. 23, 96–99 (1962)
Kayalar, S., Weinert, H.L.: Error bounds for the method of alternating projections. Math. Control Signals Systems. 1, 43–59 (1988)
Luna-Elizarrarás, M.E., Ramírez-Reyes, F., Shapiro, M.: Complexifications of real spaces general aspects. Georgian Math. J. 19, 259–282 (2012)
Von Neumann, J.: On rings of operators. Reduction theory. Ann of Math. 50(2), 401–485 (1949)
Pierra, G.: Decomposition through formalization in a product space. Math. Programming. 28, 96–115 (1984)
Popa, C. Projection algorithms: classical results and developments, applications to image reconstruction. Lambert Academic Publishing (2012)
Reich, S., Zalas, R.: The optimal error bound for the method of simultaneous projections. J. Approx. Theory. 223, 96–107 (2017)
Reich, S,. Zalas, R. Error bounds for the method of simultaneous projections with infinitely many subspaces. J. Approx. Theory. 272, 24 pp., Paper No. 105648 (2021)
Simon, B.: Real analysis: a comprehensive course in analysis, Part 1. American Mathematical Society. Providence. RI, With a 68 page companion booklet, (2015)
Acknowledgements
Both authors are grateful to two anonymous referees for their pertinent comments and helpful suggestions.
Funding
This work was partially supported by the Israel Science Foundation (Grants 389/12 and 820/17), the Fund for the Promotion of Research at the Technion and by the Technion General Research Fund.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this section we sketch how to derive Theorems 1.4 and 1.5 in a real Hilbert space, having in mind that the corresponding results of [4] were established in a complex Hilbert space. We also present an alternative proof of Theorem 1.4 by using [16, Lemma 5.2]. For this purpose, we use a complexification argument. For more details concerning the complexification, we refer the reader to [26].
To this end, let \( \textbf{H}_{\mathbb C}:= \mathcal {H} + i\mathcal {H}\) be the (external) complexification of \(\mathcal H\) with scalar multiplication given by
and inner product \(\langle \cdot , \cdot \rangle _{\mathbb C}\) defined by
where \(\alpha ,\beta \in \mathbb R \) and \(x,y,x',y'\in \mathcal {H}\). Thus, the induced norm on \( \textbf{H}_{\mathbb C} \), denoted by \(\Vert \cdot \Vert _{\mathbb C}\), satisfies
for all \(x+iy\in \textbf{H}_{\mathbb C} \). It is not difficult to see that \(( \textbf{H}_{\mathbb C}, \langle \cdot , \cdot \rangle _{\mathbb C})\) is indeed a complex Hilbert space.
Proof of Theorem 1.4
For each \(j = 1, \ldots , m\), let \(\textbf{M}_j:= M_j + i M_j\). Observe that \(\textbf{M}_j \) is a closed linear subspace of \( \textbf{H}_{\mathbb C} \). Denote by \(P_{\textbf{M}_j }\) the orthogonal projection onto \(\textbf{M}_j \). Then, for each \( \textbf{z}= x+iy \in \textbf{H}_{\mathbb C} \), we have \( P_{\textbf{M}_j}(\textbf{z}) = P_{M_j}(x) + i P_{M_j}(y)\). This implies that the product \(\textbf{T}:= P_{\textbf{M}_m} \ldots P_{\textbf{M}_1} \) satisfies \( \textbf{T}( \textbf{z}) = T(x) +iT(y)\), where T is defined as in (1.2). Using induction, we get
By [4, Remark 4.2(b) and Theorem 2.1], we have
for all \( \textbf{z}\in \textbf{H}_{\mathbb C} \). In particular, by taking \( \textbf{z}:= y_0 +i0\), we obtain
This implies Theorem 1.4.\(\square \)
Alternative proof of Theorem 1.4
By [16, Lemma 5.2] applied to the operator \( \textbf{T}\), we have
for all \( \textbf{z}\in \textbf{H}_{\mathbb C}\). In particular, by taking \( \textbf{z}:= y_0 +i0\), and knowing that the sequence \(\{\Vert y_k - y_{k-1}\Vert \}_{k=1}^\infty \) is decreasing, we have
as \(k \rightarrow \infty \). Thus we have shown that \(\Vert y_k-y_{k-1}\Vert = o(k^{-1})\) for all \(y_0\in \mathcal H\).\(\square \)
Proof of Theorem 1.5
It is not difficult to see that \(\textbf{M}_j^\perp = M_j^\perp + i M_j^\perp \) for all \(j = 1,\ldots ,m\); compare with (2.5). Consequently,
and thus
Consider now the subspaces of \(\textbf{H}_{\mathbb C}\) given by \(\textbf{X}_p:= \textbf{M}\oplus (\textbf{I}- \textbf{T})^p(\textbf{H}_{\mathbb C})\) and \(\textbf{X}:= \textstyle \bigcap _{p=1}^\infty \textbf{X}_p\), where \(\textbf{M}:= M+iM\) and \(\textbf{I}\) is the identity operator on \(\textbf{H}_{\mathbb C}\), \(p=1,2,\ldots \). Obviously, \(\textbf{X}_p\) and \(\textbf{X}\) are the analogues of \(X_p\) and X considered in Theorem 1.5. In fact, we have
Consequently, we obtain
By [4, Theorem 4.3], for each \(\textbf{z}\in \textbf{X}_p\), we get
where \(p = 1,2,\ldots \). Thus, for each \(y_0 \in X_p\) it suffices to take \(\textbf{z}:= y_0 +i0 \in \textbf{X}_p\), to see that
which shows (1.8). Similarly, for each \(y_0 \in X\), it suffices to take \(\textbf{z}:= y_0 +i0 \in \textbf{X}\) to see that (4.87) holds for all \(p > 0\). Moreover, by [4, Theorem 4.3], we know that \(\textbf{X}_p\) and \(\textbf{X}\) are dense in \(\textbf{H}_{\mathbb C}\).\(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Reich, S., Zalas, R. Polynomial estimates for the method of cyclic projections in Hilbert spaces. Numer Algor 94, 1217–1242 (2023). https://doi.org/10.1007/s11075-023-01533-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11075-023-01533-w