Spectral Discovery of Jointly Smooth Features for Multimodal Data

In this paper, we propose a spectral method for deriving functions that are jointly smooth on multiple observed manifolds. Our method is unsupervised and primarily consists of two steps. First, using kernels, we obtain a subspace spanning smooth functions on each manifold. Then, we apply a spectral method to the obtained subspaces and discover functions that are jointly smooth on all manifolds. We show analytically that our method is guaranteed to provide a set of orthogonal functions that are as jointly smooth as possible, ordered from the smoothest to the least smooth. In addition, we show that the proposed method can be efficiently extended to unseen data using the Nystrom method. We demonstrate the proposed method on both simulated and real measured data and compare the results to nonlinear variants of the seminal Canonical Correlation Analysis (CCA). Particularly, we show superior results for sleep stage identification. In addition, we show how the proposed method can be leveraged for finding minimal realizations of parameter spaces of nonlinear dynamical systems.

[1]  Maks Ovsjanikov,et al.  Functional maps , 2012, ACM Trans. Graph..

[2]  B. O. Koopman,et al.  Hamiltonian Systems and Transformation in Hilbert Space. , 1931, Proceedings of the National Academy of Sciences of the United States of America.

[3]  I. Johnstone.  MULTIVARIATE ANALYSIS AND JACOBI ENSEMBLES: LARGEST EIGENVALUE, TRACY-WIDOM LIMITS AND RATES OF CONVERGENCE. , 2008, Annals of statistics.

[4]  S. Rosenberg.  The Laplacian on a Riemannian Manifold: The Construction of the Heat Kernel , 1997 .

[5]  Jeffrey M. Hausdorff,et al.  Physionet: Components of a New Research Resource for Complex Physiologic Signals". Circu-lation Vol , 2000 .

[6]  Stéphane Lafon,et al.  Diffusion maps , 2006 .

[7]  Shotaro Akaho,et al.  A kernel method for canonical correlation analysis , 2006, ArXiv.

[8]  Ioannis G. Kevrekidis,et al.  Manifold learning for parameter reduction , 2018, J. Comput. Phys..

[9]  Mikhail Belkin,et al.  Laplacian Eigenmaps and Spectral Techniques for Embedding and Clustering , 2001, NIPS.

[10]  H. Hotelling.  Relations Between Two Sets of Variates , 1936 .

[11]  Roy R. Lederman,et al.  Learning the geometry of common latent variables using alternating-diffusion , 2015 .

[12]  Nicolas Le Roux,et al.  Out-of-Sample Extensions for LLE, Isomap, MDS, Eigenmaps, and Spectral Clustering , 2003, NIPS.

[13]  Colin Fyfe,et al.  Kernel and Nonlinear Canonical Correlation Analysis , 2000, IJCNN.

[14]  Bart De Moor,et al.  Kernel-based Data Fusion for Machine Learning - Methods and Applications in Bioinformatics and Text Mining , 2009, Studies in Computational Intelligence.

[15]  D. Giannakis.  Data-driven spectral decomposition and forecasting of ergodic dynamical systems , 2015, Applied and Computational Harmonic Analysis.

[16]  M Hirshkowitz,et al.  Atlas, rules, and recording techniques for the scoring of cyclic alternating pattern (CAP) in human sleep. , 2001, Sleep medicine.

[17]  Jeff A. Bilmes,et al.  Deep Canonical Correlation Analysis , 2013, ICML.

[18]  Bernhard Schölkopf,et al.  Kernel Principal Component Analysis , 1997, ICANN.

[19]  R. Coifman,et al.  Non-linear independent component analysis with diffusion maps , 2008 .

[20]  Karen Livescu,et al.  Nonparametric Canonical Correlation Analysis , 2015, ICML.

[21]  Jitendra Malik,et al.  Efficient spatiotemporal grouping using the Nystrom method , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[22]  John D. Storey,et al.  Statistical significance of variables driving systematic variation in high-dimensional data , 2013, Bioinform..

[23]  J. Mercer.  Functions of Positive and Negative Type, and their Connection with the Theory of Integral Equations , 1909 .

[24]  Davide Eynard,et al.  Multimodal Manifold Analysis by Simultaneous Diagonalization of Laplacians , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[25]  Hyunjoong Kim,et al.  Functional Analysis I , 2017 .

[26]  Y. Aflalo,et al.  Best bases for signal spaces , 2016 .

[27]  R. Coifman,et al.  Geometric harmonics: A novel tool for multiscale out-of-sample extension of empirical functions , 2006 .

[28]  Ron Kimmel,et al.  On the Optimality of Shape and Data Representation in the Spectral Domain , 2014, SIAM J. Imaging Sci..

[29]  I. Mezić,et al.  Applied Koopmanism. , 2012, Chaos.

[30]  Clarence W. Rowley,et al.  A Data–Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition , 2014, Journal of Nonlinear Science.

[31]  Matthias W. Seeger,et al.  Using the Nyström Method to Speed Up Kernel Machines , 2000, NIPS.

[32]  Hans-Joachim Bungartz,et al.  Numerical Model Construction with Closed Observables , 2015, SIAM J. Appl. Dyn. Syst..

[33]  Haim Brezis,et al.  Rigidity of optimal bases for signal spaces , 2017 .

[34]  Nello Cristianini,et al.  Kernel Methods for Pattern Analysis , 2003, ICTAI.

[35]  Christopher R. Myers,et al.  Universally Sloppy Parameter Sensitivities in Systems Biology Models , 2007, PLoS Comput. Biol..