Abstract
Let \({\mathsf {S}}_1\) (the Schatten–von Neumann trace class) denote the Banach space of all compact linear operators \(T:\ell _2\rightarrow \ell _2\) whose nuclear norm \(\Vert T\Vert _{{\mathsf {S}}_1}=\sum _{j=1}^\infty \upsigma _j(T)\) is finite, where \(\{\upsigma _j(T)\}_{j=1}^\infty \) are the singular values of T. We prove that for arbitrarily large \(n\in {\mathbb {N}}\) there exists a subset \({\mathscr {C}}\subseteq {\mathsf {S}}_1\) with \(|{\mathscr {C}}|=n\) that cannot be embedded with bi-Lipschitz distortion O(1) into any \(n^{o(1)}\)-dimensional linear subspace of \({\mathsf {S}}_1\). \({\mathscr {C}}\) is not even a O(1)-Lipschitz quotient of any subset of any \(n^{o(1)}\)-dimensional linear subspace of \({\mathsf {S}}_1\). Thus, \({\mathsf {S}}_1\) does not admit a dimension reduction result á la Johnson and Lindenstrauss (Conference in modern analysis and probability, American Mathematical Society, Providence, 1984), which complements the work of Harrow et al. (Automata, languages and programming (ICALP 2011), Springer, Heidelberg, 2011) on the limitations of quantum dimension reduction under the assumption that the embedding into low dimensions is a quantum channel. Such a statement was previously known with \({\mathsf {S}}_1\) replaced by the Banach space \(\ell _1\) of absolutely summable sequences via the work of Brinkman and Charikar (J ACM 52(5):766–788, 2005). In fact, the above set \({\mathscr {C}}\) can be taken to be the same set as the one that Brinkman and Charikar considered, viewed as a collection of diagonal matrices in \({\mathsf {S}}_1\). The challenge is to demonstrate that \({\mathscr {C}}\) cannot be faithfully realized in an arbitrary low-dimensional subspace of \({\mathsf {S}}_1\), while Brinkman and Charikar obtained such an assertion only for subspaces of \({\mathsf {S}}_1\) that consist of diagonal operators (i.e., subspaces of \(\ell _1\)). We establish this by proving that the Markov 2-convexity constant of any finite dimensional linear subspace X of \({\mathsf {S}}_1\) is at most a universal constant multiple of \(\sqrt{\log \dim (X)}\).
Similar content being viewed by others
Notes
Formally, for the purpose of efficient approximate nearest neighbor search one cannot use a dimension reduction statement like (2) as a “black box” without additional information about the low-dimensional embedding itself rather than its mere existence. One would want the embedding to be fast to compute and “data oblivious,” as in the classical Johnson–Lindenstrauss lemma [45]. There is no need to give a precise formulation here because the present article is devoted to ruling out any low-dimensional low-distortion embedding whatsoever.
We shall use throughout this article the following (standard) asymptotic notation. Given two quantities \(Q,Q'>0\), the notations \(Q\lesssim Q'\) and \(Q'\gtrsim Q\) mean that \(Q\leqslant {\mathsf {K}}Q'\) for some universal constant \({\mathsf {K}}>0\). The notation \(Q\asymp Q'\) stands for \((Q\lesssim Q') \wedge (Q'\lesssim Q)\). If we need to allow for dependence on certain parameters, we indicate this by subscripts. For example, in the presence of an auxiliary parameter \(\uppsi \), the notation \(Q\lesssim _\uppsi Q'\) means that \(Q\leqslant c(\uppsi )Q' \), where \(c(\uppsi ) >0\) is allowed to depend only on \(\uppsi \), and similarly for the notations \(Q\gtrsim _\uppsi Q'\) and \(Q\asymp _\uppsi Q'\).
Those who prefer to consider the nuclear norm on \(m\times m\) matrices can do so throughout, since all of our results are equivalent to their matricial counterparts; see Lemma 2.1 below for a formulation of this (straightforward) statement.
One should note here that \({\mathsf {S}}_1\) does not admit a bi-Lipschitz embedding into any \(L_1(\upmu )\) space, as follows by combining the corresponding linear result of [66, 85] with a classical differentiation argument [12], or directly by using a bi-Lipschitz invariant that is introduced in the forthcoming work [79].
For many purposes it is important to work with complex scalars, and correspondingly complex matrices. However, for the purpose of the ensuing metric results, statements over \({\mathbb {R}}\) are equivalent to their complex counterparts.
References
Alon, N.: Problems and results in extremal combinatorics. I. Discrete Math. 273(1–3), 31–53 (2003)
Andoni, A.: Nearest neighbor search in high-dimensional spaces. In: Murlak F., Sankowski P. (eds.) Proceedings of the 36th International Symposium on Mathematical Foundations of Computer Science (MFCS 2011). Lecture Notes in Computer Science, vol. 6907, pp. 1. Springer, Berlin (2011). www.mit.edu/~andoni/papers/nns-mfcs.pptx
Andoni, A., Charikar, M.S., Neiman, O., Nguy\(\tilde{\hat{{\rm e}}}\)n, H.L.: Near linear lower bound for dimension reduction in \(\ell _1\). In: 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science (FOCS 2011), pp. 315–323. IEEE Computer Society, Los Alamitos (2011)
Andoni, A., Naor, A., Neiman, O.: On isomorphic dimension reduction in \(\ell _1\) (2017). Preprint
Andoni, A., Naor, A., Nikolov, A., Razenshteyn, I., Waingarten, E.: Data-dependent hashing via nonlinear spectral gaps. In: STOC’18—Proceedings of the 50th Annual ACM SIGACT Symposium on Theory of Computing, pp. 787–800. ACM, New York (2018)
Artstein, S.: Proportional concentration phenomena on the sphere. Isr. J. Math. 132, 337–358 (2002)
Ball, K.: The Ribe programme. Séminaire Bourbaki. Vol. 2011/2012. Exposés 1043–1058. Astérisque No. 352, Exp. No. 1047, 147–159 (2013)
Ball, K.: Markov chains, Riesz transforms and Lipschitz maps. Geom. Funct. Anal. 2(2), 137–172 (1992)
Ball, K., Carlen, E.A., Lieb, E.H.: Sharp uniform convexity and smoothness inequalities for trace norms. Invent. Math. 115(3), 463–482 (1994)
Bartal, Y., Linial, N., Mendel, M., Naor, A.: On metric Ramsey-type phenomena. Ann. Math. 162(2), 643–709 (2005)
Bates, S., Johnson, W.B., Lindenstrauss, J., Preiss, D., Schechtman, G.: Affine approximation of Lipschitz functions and nonlinear quotients. Geom. Funct. Anal. 9(6), 1092–1127 (1999)
Benyamini, Y., Lindenstrauss, J.: Geometric Nonlinear Functional Analysis, vol. 1. American Mathematical Society Colloquium Publications, vol. 48. American Mathematical Society, Providence (2000)
Bhatia, R.: Matrix Analysis. Graduate Texts in Mathematics, vol. 169. Springer, New York (1997)
Bourgain, J.: On Lipschitz embedding of finite metric spaces in Hilbert space. Isr. J. Math. 52(1–2), 46–52 (1985)
Bourgain, J., Lindenstrauss, J., Milman, V.: Approximation of zonoids by zonotopes. Acta Math. 162(1–2), 73–141 (1989)
Bourgain, J., Milman, V., Wolfson, H.: On type of metric spaces. Trans. Amer. Math. Soc. 294(1), 295–317 (1986)
Bouwmans, T., Serhat-Aybat, N., Zahzah, E. (eds.): Handbook of Robust Low-Rank and Sparse Matrix Decomposition: Applications in Image and Video Processing. CRC Press, Boca Raton (2016)
Brinkman, B., Charikar, M.: On the impossibility of dimension reduction in \(l_1\). J. ACM 52(5), 766–788 (2005)
Candès, E.J., Recht, B.: Exact matrix completion via convex optimization. Found. Comput. Math. 9(6), 717–772 (2009)
Candès, E.J., Tao, T.: The power of convex relaxation: near-optimal matrix completion. IEEE Trans. Inf. Theory 56(5), 2053–2080 (2010)
Carlen, E.: Trace Inequalities and Quantum Entropy: An Introductory Course. Entropy and the Quantum. Contemporary Mathematics, vol. 529, pp. 73–140. American Mathematical Society, Providence (2010)
Charikar, M., Sahai, A.: Dimension reduction in the \(\ell _1\) norm. In: Proceedings of the 43rd Symposium on Foundations of Computer Science (FOCS 2002), pp. 551–560. IEEE Computer Society (2002)
Clarkson, J.A.: Uniformly convex spaces. Trans. Amer. Math. Soc. 40(3), 396–414 (1936)
Deshpande, A., Tulsiani, M., Vishnoi, N.K.: Algorithms and hardness for subspace approximation. In: Proceedings of the 22nd Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 482–496. SIAM, Philadelphia (2011)
Dilworth, S.J.: On the dimension of almost Hilbertian subspaces of quotient spaces. J. London Math. Soc. 30(3), 481–485 (1984)
Dixmier, J.: Formes linéaires sur un anneau d’opérateurs. Bull. Soc. Math. France 81, 9–39 (1953)
Dixmier, J.: Les algèbres d’opérateurs dans l’espace hilbertien (algèbres de von Neumann). Les Grands Classiques Gauthier-Villars. Éditions Jacques Gabay, Paris. Reprint of the second 1969 edition (1996)
Dvoretzky, A.: Some results on convex bodies and Banach spaces. In: Proceedings of the International Symposium on Linear Spaces, pp. 123–160. Jerusalem Academic Press, Jerusalem (1961)
Enflo, P.: On the nonexistence of uniform homeomorphisms between \(L_{p}\)-spaces. Ark. Mat. 8, 103–105 (1969)
Enflo, P.: On infinite-dimensional topological groups. In: Séminaire sur la Géométrie des Espaces de Banach (1977–1978), Exp. No. 10–11. École Polytechnique, Palaiseau (1978)
Eskenazis, A., Mendel, M., Naor, A.: Diamond convexity: a bifurcation in the Ribe program (2017). Preprint
Figiel, T., Lindenstrauss, J., Milman, V.D.: The dimension of almost spherical sections of convex bodies. Acta Math. 139(1–2), 53–94 (1977)
Gromov, M.: Metric Structures for Riemannian and Non-Riemannian Spaces. Progress in Mathematics, vol. 152. Birkhäuser, Boston (1999)
Gu, S., Xie, Q., Meng, D., Zuo, W., Feng, X., Zhang, L.: Weighted nuclear norm minimization and its applications to low level vision. Int. J. Comput. Vis. 121(2), 183–208 (2017)
Gu, S., Zhang, L., Zuo, W., Feng, X.: Weighted nuclear norm minimization with application to image denoising. In: 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2014), pp. 2862–2869. IEEE Computer Society (2014)
Gupta, A., Newman, I., Rabinovich, Yu., Sinclair, A.: Cuts, trees and \(l_1\)-embeddings of graphs. Combinatorica 24(2), 233–269 (2004)
Gutman, I.: The energy of a graph. In: 10. Steiermärkisches Mathematisches Symposium (Stift Rein, Graz, 1978). Ber. Math.-Statist. Sekt. Forsch. Graz (100–105), Ber. No. 103 (1978)
Harchaoui, Z., Douze, M., Paulin, M., Dudík, M., Malick, J.: Large-scale image classification with trace-norm regularization. In: Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition, pp. 3386–3393. IEEE Computer Society (2012)
Hardt, M., Ligett, K., McSherry, F.: A simple and practical algorithm for differentially private data release. In: Bartlett, P.L., et al. (eds.) Advances in Neural Information Processing Systems, vol. 2, pp. 2348–2356. Curran Associates, New York (2012)
Harrow, A.W., Montanaro, A., Short, A.J.: Limitations on quantum dimensionality reduction. In: Aceto, L., Henzinger, M., Sgall, J. (eds.) Automata, Languages and Programming (ICALP 2011). Part I. Lecture Notes in Computer Science, vol. 6755, pp. 86–97. Springer, Heidelberg (2011)
Heinonen, J.: Lectures on Analysis on Metric Spaces. Universitext. Springer, New York (2001)
Hytönen, T., Naor, A.: Heat flow and quantitative differentiation. J. Eur. Math. Soc. (JEMS) 21(11), 3415–3466 (2019)
Indyk, P., Naor, A.: Nearest-neighbor-preserving embeddings. ACM Trans. Algorithms 3(3), 31 (2007)
James, I.M.: Introduction to Uniform Spaces. London Mathematical Society Lecture Note Series, vol. 144. Cambridge University Press, Cambridge (1990)
Johnson, W.B., Lindenstrauss, J.: Extensions of Lipschitz mappings into a Hilbert space. In: Beals, A., et al. (eds.) Conference in Modern Analysis and Probability. Contemporary Mathematics, vol. 26, pp. 189–206. American Mathematical Society, Providence (1984)
Johnson, W.B., Lindenstrauss, J., Schechtman, G.: On Lipschitz embedding of finite metric spaces in low-dimensional normed spaces. In: Lindenstrauss, J., Milman, V.D. (eds.) Geometrical Aspects of Functional Analysis (1985/86). Lecture Notes in Mathematics, vol. 1267, pp. 177–184. Springer, Berlin (1987)
Johnson, W.B., Naor, A.: The Johnson–Lindenstrauss lemma almost characterizes Hilbert space, but not quite. Discrete Comput. Geom. 43(3), 542–553 (2010)
Johnson, W.B., Schechtman, G.: Diamond graphs and super-reflexivity. J. Topol. Anal. 1(2), 177–189 (2009)
Laakso, T.J.: Ahlfors \(Q\)-regular spaces with arbitrary \(Q>1\) admitting weak Poincaré inequality. Geom. Funct. Anal. 10(1), 111–123 (2000)
Lafforgue, V., Naor, A.: Vertical versus horizontal Poincaré inequalities on the Heisenberg group. Isr. J. Math. 203(1), 309–339 (2014)
Lang, U., Plaut, C.: Bilipschitz embeddings of metric spaces into space forms. Geom. Dedicata 87(1–3), 285–307 (2001)
Larsen, K.G., Nelson, J.: Optimality of the Johnson–Lindenstrauss lemma. In: Proceedings of the 58th Annual IEEE Symposium on Foundations of Computer Science (FOCS 2017), pp. 633–638. IEEE Computer Society, Los Alamitos (2017). https://arxiv.org/abs/1609.02094
Lee, J.R., Naor, A.: Embedding the diamond graph in \(L_p\) and dimension reduction in \(L_1\). Geom. Funct. Anal. 14(4), 745–747 (2004)
Lee, J.R., Naor, A., Peres, Y.: Trees and Markov convexity. Geom. Funct. Anal. 18(5), 1609–1659 (2009)
Lee, J.R., Mendel, M., Naor, A.: Metric structures in \(L_1\): dimension, snowflakes, and average distortion. Eur. J. Comb. 26(8), 1180–1190 (2005)
Lewis, D.R.: Finite dimensional subspaces of \(L_{p}\). Stud. Math. 63(2), 207–212 (1978)
Li, C., Miklau, G.: Optimal error of query sets under the differentially-private matrix mechanism. In: Tan, W., et al. (eds.) Proceedings 16th International Conference on Database Theory (ICDT’13), pp. 272–283. ACM, New York (2013)
Li, X., Shi, Y., Gutman, I.: Graph Energy. Springer, New York (2012)
Li, Y., Woodruff, D.P.: Embeddings of Schatten norms with applications to data streams (2017). To appear in Proceedings of 44th International Colloquium on Automata, Languages, and Programming, ICALP 2017. https://arxiv.org/abs/1702.05626
Li, Y., Woodruff, D.P.: Embeddings of Schatten norms with applications to data streams. In: Proceedings of the 44th International Colloquium on Automata, Languages, and Programming (ICALP 2017). LIPIcs. Leibniz International Proceedings in Informatics, vol. 80, Art. No. 60. Schloss Dagstuhl. Leibniz-Zentrum für Informatik, Wadern (2017)
Linial, N., London, E., Rabinovich, Yu.: The geometry of graphs and some of its algorithmic applications. Combinatorica 15(2), 215–245 (1995)
Löwner, K.: Über monotone Matrixfunktionen. Math. Z. 38(1), 177–216 (1934)
Martínez, T., Torrea, J.L., Xu, Q.: Vector-valued Littlewood–Paley–Stein theory for semigroups. Adv. Math. 203(2), 430–475 (2006)
Matoušek, J.: Note on bi-Lipschitz embeddings into normed spaces. Comment. Math. Univ. Carolin. 33(1), 51–55 (1992)
Matoušek, J.: On the distortion required for embedding finite metric spaces into normed spaces. Isr. J. Math. 93, 333–344 (1996)
McCarthy, C.A.: \(c_{p}\). Isr. J. Math. 5, 249–271 (1967)
Mendel, M., Naor, A.: Euclidean quotients of finite metric spaces. Adv. Math. 189(2), 451–494 (2004)
Mendel, M., Naor, A.: Markov convexity and local rigidity of distorted metrics. J. Eur. Math. Soc. (JEMS) 15(1), 287–337 (2013)
Mendel, M., Naor, A.: Spectral calculus and Lipschitz extension for barycentric metric spaces. Anal. Geom. Metr. Spaces 1, 163–199 (2013)
Mendel, M., Naor, A.: Nonlinear spectral calculus and super-expanders. Publ. Math. Inst. Hautes Étud. Sci. 119, 1–95 (2014)
Milman, V.D.: A new proof of A. Dvoretzky’s theorem on cross-sections of convex bodies. Funkcional Anal. Priložen. 5(4), 28–37 (1971)
Milman, V.D.: Almost Euclidean quotient spaces of subspaces of a finite-dimensional normed space. Proc. Am. Math. Soc. 94(3), 445–449 (1985)
Mirsky, L.: A trace inequality of John von Neumann. Monatsh. Math. 79(4), 303–306 (1975)
Naor, A.: An introduction to the Ribe program. Jpn. J. Math. 7(2), 167–233 (2012)
Naor, A.: Discrete Riesz transforms and sharp metric \(X_p\) inequalities. Ann. Math. 184(3), 991–1016 (2016)
Naor, A.: A spectral gap precludes low-dimensional embeddings. In: Aronov, B., Katz, M.J. (eds.) 33rd International Symposium on Computational Geometry (SoCG 2017). Leibniz International Proceedings in Informatics, vol. 77, pp. 50:1–50:16. Schloss Dagstuhl. Leibniz-Zentrum für Informatik, Wadern (2017)
Naor, A.: Metric dimension reduction: a snapshot of the Ribe program. In: Proceedings of the 2018 International Congress of Mathematicians, Rio de Janeiro, vol. I, pp. 767–846 (2018)
Naor, A., Schechtman, G.: Metric \(X_p\) inequalities. Forum Math. Pi 4, e3 (2016)
Naor, A., Schechtman, G.: Obstructions to metric embeddings of Schatten classes (2017). Preprint
Naor, A., Young, R.: Vertical perimeter versus horizontal perimeter. Ann. Math. 188(1), 171–279 (2018)
Naor, A., Young, R.: Foliated corona decompositions (2019). Preprint
Nikiforov, V.: The energy of graphs and matrices. J. Math. Anal. Appl. 326(2), 1472–1475 (2007)
Ohta, S.-I.: Markov type of Alexandrov spaces of non-negative curvature. Mathematika 55(1–2), 177–189 (2009)
Pisier, G.: Martingales with values in uniformly convex spaces. Isr. J. Math. 20(3–4), 326–350 (1975)
Pisier, G.: Some results on Banach spaces without local unconditional structure. Compos. Math. 37(1), 3–19 (1978)
Pisier, G.: Probabilistic methods in the geometry of Banach spaces. In: Letta, G., Pratelli, M. (eds.) Probability and Analysis. Lecture Notes in Mathematics, vol. 1206, pp. 167–241. Springer, Berlin (1986)
Recht, B.: A simpler approach to matrix completion. J. Mach. Learn. Res. 12, 3413–3430 (2011)
Recht, B., Fazel, M., Parrilo, P.A.: Guaranteed minimum-rank solutions of linear matrix equations via nuclear norm minimization. SIAM Rev. 52(3), 471–501 (2010)
Regev, O.: Entropy-based bounds on dimension reduction in \(L_1\). Isr J. Math. 195(2), 825–832 (2013)
Regev, O., Vidick, T.: Bounds on dimension reduction in the nuclear norm (2019). https://arxiv.org/abs/1901.09480
Schechtman, G.: More on embedding subspaces of \(L_p\) in \(l^n_r\). Compos. Math. 61(2), 159–169 (1987)
Schechtman, G., Zvavitch, A.: Embedding subspaces of \(L_p\) into \(l^N_p\), \(0<p<1\). Math. Nachr. 227, 133–142 (2001)
Simon, B.: Trace Ideals and Their Applications. Mathematical Surveys and Monographs, vol. 120, 2nd edn. American Mathematical Society, Providence (2005)
Talagrand, M.: Embedding subspaces of \(L_1\) into \(l^N_1\). Proc. Am. Math. Soc. 108(2), 363–369 (1990)
Tomczak-Jaegermann, N.: Finite-dimensional subspaces of uniformly convex and uniformly smooth Banach lattices and trace classes \(S_{p}\). Stud. Math. 66(3), 261–281 (1979–1980)
von Neumann, J.: Some matrix-inequalities and metrization of matric-space. Tomsk Univ. Rev. 1, 286–300 (1937). Reprinted in Collected Works, Pergamon Press (1962), iv, pp. 205–219
Whyburn, G.T.: Topological Analysis. Princeton Mathematical Series, vol. 23. Princeton University Press, Princeton (1958)
Winter, A.: Quantum and classical message protect identification via quantum channels. Quantum Inf. Comput. 4(6–7), 563–578 (2004)
Xu, Q.: Littlewood–Paley theory for functions with values in uniformly convex spaces. J. Reine Angew. Math. 504, 195–226 (1998)
Acknowledgements
We thank A. Andoni, R. Krauthgamer and M. Mendel for helpful input.
Author information
Authors and Affiliations
Corresponding author
Additional information
Editor in Charge: Kenneth Clarkson
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
A. Naor was supported by the BSF, the NSF, the Packard Foundation and the Simons Foundation. G. Schechtman was supported by the ISF. The research that is presented here was conducted under the auspices of the Simons Algorithms and Geometry (A&G) Think Tank. An extended abstract that announces the results that are obtained here appeared in the proceedings of the 29th annual ACM–SIAM Symposium on Discrete Algorithms (SODA 2018).
Rights and permissions
About this article
Cite this article
Naor, A., Pisier, G. & Schechtman, G. Impossibility of Dimension Reduction in the Nuclear Norm. Discrete Comput Geom 63, 319–345 (2020). https://doi.org/10.1007/s00454-019-00162-2
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00454-019-00162-2
Keywords
- Dimension reduction
- Metric embeddings
- Nuclear norm
- Schatten–von Neumann classes
- Lipschitz quotients
- Markov convexity