Towards a Mathematical Theory of Super-resolution

Size: px
Start display at page:

Download "Towards a Mathematical Theory of Super-resolution"

Transcription

1 Towards a Mathematical Theory of Super-resolution Carlos Fernandez-Granda Information Theory Forum, Information Systems Laboratory, Stanford 10/18/2013

2 Acknowledgements This work was supported by a Fundación La Caixa Fellowship and a Fundación Caja Madrid Fellowship Collaborator : Emmanuel Candès (Department of Mathematics and of Statistics, Stanford)

3 Motivation : Limits of resolution in imaging The resolving power of lenses, however perfect, is limited (Lord Rayleigh) Diraction imposes a fundamental limit on the resolution of optical systems

4 Motivation Similar problems arise in electronic imaging, signal processing, radar, spectroscopy, medical imaging, astronomy, geophysics, etc.

5 Motivation Similar problems arise in electronic imaging, signal processing, radar, spectroscopy, medical imaging, astronomy, geophysics, etc. Signals of interest are often point sources : celestial bodies (astronomy), line spectra (signal processing), molecules (uorescence microscopy),...

6 Super-resolution Aim : estimating ne-scale structure from low-resolution data

7 Super-resolution Aim : estimating ne-scale structure from low-resolution data Equivalently, extrapolating the high end of the spectrum

8 Mathematical model Signal : superposition of Dirac measures with support T x = j a j δ tj a j C, t j T [0, 1]

9 Mathematical model Signal : superposition of Dirac measures with support T x = j a j δ tj a j C, t j T [0, 1] Measurements : low-pass ltering with cut-o frequency f c y = F c x y(k) = 1 0 (vector of low-pass Fourier coecients) e i 2πkt x (dt) = a j e i 2πkt j, k Z, k f c j

10 Equivalent problem : line-spectra estimation Swapping time and frequency Signal : superposition of sinusoids x(t) = j a j e i 2πf j t a j C, f j T [0, 1]

11 Equivalent problem : line-spectra estimation Swapping time and frequency Signal : superposition of sinusoids x(t) = j a j e i 2πf j t a j C, f j T [0, 1] Measurements : equispaced samples x(1), x(2), x(3),... x(n)

12 Equivalent problem : line-spectra estimation Swapping time and frequency Signal : superposition of sinusoids x(t) = j a j e i 2πf j t a j C, f j T [0, 1] Measurements : equispaced samples x(1), x(2), x(3),... x(n) Classical problem in signal processing

13 Can you nd the spikes?

14 Can you nd the spikes?

15 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,...

16 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed

17 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D)

18 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D) Noise must be Gaussian and white

19 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D) Noise must be Gaussian and white Incorporating additional structure is dicult (dierent point-spread functions, assumptions on the signal)

20 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D) Noise must be Gaussian and white Incorporating additional structure is dicult (dierent point-spread functions, assumptions on the signal) This talk Non-parametric estimation

21 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D) Noise must be Gaussian and white Incorporating additional structure is dicult (dierent point-spread functions, assumptions on the signal) This talk Non-parametric estimation Provably stable in the presence of noise

22 Prior art Based on Prony's method : MUSIC, ESPRIT, matrix pencil,... An estimate of the number of spikes is needed Do not apply to multiple dimensions (not even 2D) Noise must be Gaussian and white Incorporating additional structure is dicult (dierent point-spread functions, assumptions on the signal) This talk Non-parametric estimation Provably stable in the presence of noise Flexible variational framework based on convex programming

23 Outline of the talk Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

24 Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

25 Compressed sensing vs super-resolution Estimation of sparse signals from undersampled measurements suggests connections to compressed sensing

26 Compressed sensing vs super-resolution Estimation of sparse signals from undersampled measurements suggests connections to compressed sensing Compressed sensing Super-resolution spectrum interpolation spectrum extrapolation

27 Compressed sensing vs super-resolution Compressed sensing Super-resolution spectrum interpolation spectrum extrapolation

28 Compressed sensing Compressed sensing theory establishes robust recovery of spikes from random Fourier measurements [Candès, Romberg & Tao 2004]

29 Compressed sensing Compressed sensing theory establishes robust recovery of spikes from random Fourier measurements [Candès, Romberg & Tao 2004] Crucial insight : measurement operator is well conditioned when acting upon sparse signals

30 Compressed sensing Compressed sensing theory establishes robust recovery of spikes from random Fourier measurements [Candès, Romberg & Tao 2004] Crucial insight : measurement operator is well conditioned when acting upon sparse signals Equivalently, the energy of all sparse signals is preserved by the randomized measurements (restricted isometry property)

31 Compressed sensing Compressed sensing theory establishes robust recovery of spikes from random Fourier measurements [Candès, Romberg & Tao 2004] Crucial insight : measurement operator is well conditioned when acting upon sparse signals Equivalently, the energy of all sparse signals is preserved by the randomized measurements (restricted isometry property) This is a necessary condition for stable estimation, but is it the case in super-resolution?

32 Simple experiment F x = y Discretize support to lie on a grid with N = 4096 points

33 Simple experiment F c x = y c Measure n low-pass DFT coecients, super-resolution factor (SRF) : N/n

34 Simple experiment F c = y c x Measure n low-pass DFT coecients, super-resolution factor (SRF) : N/n

35 Simple experiment F c,t = y c x T Restrict support of the signal to an interval of 48 contiguous points

36 Simple experiment x T = F c,t y c Compute singular values of resulting linear operator

37 Simple experiment SRF=2 SRF=4 SRF=8 SRF=16 1e 17 1e 13 1e 09 1e 05 1e SRF=2 SRF=4 SRF=8 SRF=16 For SRF = 4 measuring any unit-normed vector in a subspace of dimension 24 results in a signal with norm less than 10 7

38 Simple experiment SRF=2 SRF=4 SRF=8 SRF=16 1e 17 1e 13 1e 09 1e 05 1e SRF=2 SRF=4 SRF=8 SRF=16 For SRF = 4 measuring any unit-normed vector in a subspace of dimension 24 results in a signal with norm less than 10 7 At an SNR of 145 db, recovery is impossible by any method

39 Simple experiment SRF=2 SRF=4 SRF=8 SRF=16 1e 17 1e 13 1e 09 1e 05 1e SRF=2 SRF=4 SRF=8 SRF=16 For SRF = 4 measuring any unit-normed vector in a subspace of dimension 24 results in a signal with norm less than 10 7 At an SNR of 145 db, recovery is impossible by any method This phenomenon is characterized asymptotically by Slepian's seminal work on prolate spheroidal sequences

40 Conclusion Sparsity is not enough Signal Spectrum

41 Conclusion Sparsity is not enough Signal Spectrum Additional conditions are necessary to restrict our signal model

42 Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

43 Minimum separation To exclude highly-clustered signals from our model, we control the minimum separation of the support T = inf t t (t,t ) T : t t

44 Total-variation norm Continuous counterpart of the l 1 norm

45 Total-variation norm Continuous counterpart of the l 1 norm If x = j a jδ tj then x TV = j a j

46 Total-variation norm Continuous counterpart of the l 1 norm If x = j a jδ tj then x TV = j a j Not the total variation of a piecewise-constant function

47 Total-variation norm Continuous counterpart of the l 1 norm If x = a j jδ tj then x TV = a j j Not the total variation of a piecewise-constant function Formal denition : For a complex measure ν ν TV = sup ν (B j ), j=1 (supremum over all nite partitions B j of [0, 1])

48 Recovery via convex programming In the absence of noise, i.e. if y = F c x, we solve min x x TV subject to F c x = y, over all nite complex measures x supported on [0, 1]

49 Recovery via convex programming In the absence of noise, i.e. if y = F c x, we solve min x x TV subject to F c x = y, over all nite complex measures x supported on [0, 1] Theorem [Candès, F. 2012] If the minimum separation of the signal support T obeys 2 /f c := 2 λ c, then recovery is exact

50 Minimum-distance condition λ c /2 is the Rayleigh resolution limit (half-width of measurement lter)

51 Minimum-distance condition λ c /2 is the Rayleigh resolution limit (half-width of measurement lter) Numerical simulations show that TV-norm minimization fails if < λ c

52 Minimum-distance condition λ c /2 is the Rayleigh resolution limit (half-width of measurement lter) Numerical simulations show that TV-norm minimization fails if < λ c If < λ c /2 some signals are almost in the nullspace of the measurement operator (no method can achieve stable estimation)

53 Sparse recovery in overcomplete dictionaries If we discretize the support Sparse recovery via l 1 -norm minimization in an overcomplete Fourier dictionary

54 Sparse recovery in overcomplete dictionaries If we discretize the support Sparse recovery via l 1 -norm minimization in an overcomplete Fourier dictionary Previous theory based on dictionary incoherence is very weak, due to high column correlation

55 Sparse recovery in overcomplete dictionaries If we discretize the support Sparse recovery via l 1 -norm minimization in an overcomplete Fourier dictionary Previous theory based on dictionary incoherence is very weak, due to high column correlation If N = and n = 1000, how many spikes can we recover?

56 Sparse recovery in overcomplete dictionaries If we discretize the support Sparse recovery via l 1 -norm minimization in an overcomplete Fourier dictionary Previous theory based on dictionary incoherence is very weak, due to high column correlation If N = and n = 1000, how many spikes can we recover? Previous theory [Dossal 2005] : 3 spikes

57 Sparse recovery in overcomplete dictionaries If we discretize the support Sparse recovery via l 1 -norm minimization in an overcomplete Fourier dictionary Previous theory based on dictionary incoherence is very weak, due to high column correlation If N = and n = 1000, how many spikes can we recover? Previous theory [Dossal 2005] : 3 spikes Our result : n/4 = 250 spikes

58 Higher dimensions Signal : superposition of point sources (delta measures) in 2D Measurements : low-pass 2D Fourier coecients

59 Higher dimensions Signal : superposition of point sources (delta measures) in 2D Measurements : low-pass 2D Fourier coecients Theorem [Candès, F. 2012] TV-norm minimization yields exact recovery if 2.38 λ c

60 Higher dimensions Signal : superposition of point sources (delta measures) in 2D Measurements : low-pass 2D Fourier coecients Theorem [Candès, F. 2012] TV-norm minimization yields exact recovery if 2.38 λ c In dimension d, C d λ c, where C d only depends on d

61 Extensions Signal : piecewise-constant function Measurements : low-pass Fourier coecients

62 Extensions Signal : piecewise-constant function Measurements : low-pass Fourier coecients Corollary Solving min x (1) TV subject to F c x = y yields exact recovery if 2 λ c Similar result for cont. dierentiable piecewise-smooth functions

63 Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

64 Optimality conditions Consider the problem where f is convex Lemma min x f ( x) subject to A x = y, If there exists a subgradient g(x) of f at a feasible point x such that g(x) = A v for some v, then x is a solution

65 Optimality conditions Consider the problem where f is convex Lemma min x f ( x) subject to A x = y, If there exists a subgradient g(x) of f at a feasible point x such that g(x) = A v for some v, then x is a solution For any h such that Ah = 0 f (x + h) f (x) + g(x), h = f (x) + A v, h = f (x) + v, Ah = f (x) by denition of subgradient and convexity of f

66 Certicate of optimality q is a subgradient of the total-variation norm at x = j T a j e iφ j δ tj { q(t j ) = e iφ j, q(t) 1, t j T t [0, 1] \ T if (1)

67 Certicate of optimality q is a subgradient of the total-variation norm at x = j T a j e iφ j δ tj { q(t j ) = e iφ j, q(t) 1, t j T t [0, 1] \ T if (1) To certify optimality we also need q(t) = F c v = fc k= f c v k e i 2πkt

68 Certicate of optimality q is a subgradient of the total-variation norm at x = j T a j e iφ j δ tj { q(t j ) = e iφ j, q(t) 1, t j T t [0, 1] \ T if (1) To certify optimality we also need If (1) is strengthened to q(t) = F c v = fc k= f c v k e i 2πkt q(t) < 1, t [0, 1] \ T then x is the unique solution

69 Certicate of optimality

70 Construction of the certicate 1st idea : interpolation with a low-frequency fast-decaying kernel K q(t) = tj T α j K(t t j ),

71 Construction of the certicate 1st idea : interpolation with a low-frequency fast-decaying kernel K q(t) = tj T α j K(t t j ),

72 Construction of the certicate 1st idea : interpolation with a low-frequency fast-decaying kernel K q(t) = tj T α j K(t t j ),

73 Construction of the certicate 1st idea : interpolation with a low-frequency fast-decaying kernel K q(t) = tj T α j K(t t j ),

74 Construction of the certicate 1st idea : interpolation with a low-frequency fast-decaying kernel K q(t) = tj T α j K(t t j ),

75 Construction of the certicate Problem : magnitude of polynomial locally exceeds 1

76 Construction of the certicate Problem : magnitude of polynomial locally exceeds 1

77 Construction of the certicate Problem : magnitude of polynomial locally exceeds 1 Solution : add correction term and force q (t k ) = 0 for all t k T q(t) = tj T α j K(t t j ) + β j K (t t j )

78 Construction of the certicate Problem : magnitude of polynomial locally exceeds 1 Solution : add correction term and force q (t k ) = 0 for all t k T q(t) = tj T α j K(t t j ) + β j K (t t j )

79 Construction of the certicate Problem : magnitude of polynomial locally exceeds 1 Solution : add correction term and force q (t k ) = 0 for all t k T q(t) = tj T α j K(t t j ) + β j K (t t j )

80 Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

81 Practical implementation Primal problem : min x x TV subject to F c x = y, Innite-dimensional variable x (measure in [0, 1]) First option : Discretize domain and apply l 1 -norm minimization

82 Practical implementation Primal problem : min x x TV subject to F c x = y, Innite-dimensional variable x (measure in [0, 1]) First option : Discretize domain and apply l 1 -norm minimization Dual problem : max Re [y u] subject to F u c 1, n := 2f c + 1 u C n Finite-dimensional variable u, but innite-dimensional constraint F u = c u k e i 2πkt k f c Second option : Recast dual problem as semidenite program

83 Lemma : Semidenite representation The Fejér-Riesz Theorem and the semidenite representation of polynomial sums of squares imply that F c u 1 is equivalent to There exists a Hermitian matrix Q C n n such that [ ] Q u u 0, 1 n j Q i,i+j = i=1 { 1, j = 0, 0, j = 1, 2,..., n 1.

84 Using the dual solution We can solve the dual problem, but how do we retrieve a primal solution?

85 Using the dual solution We can solve the dual problem, but how do we retrieve a primal solution? Dual solution vector : Fourier coecients of low-pass polynomial that interpolates the sign of the primal solution (follows from strong duality)

86 Using the dual solution We can solve the dual problem, but how do we retrieve a primal solution? Dual solution vector : Fourier coecients of low-pass polynomial that interpolates the sign of the primal solution (follows from strong duality) Idea : Use the polynomial to locate the support of the signal and then estimate the amplitudes by least squares

87 Super-resolution via semidenite programming Measurements High res signal

88 Super-resolution via semidenite programming Solve semidenite program to obtain dual solution

89 Super-resolution via semidenite programming Locate points at which corresponding polynomial has unit magnitude

90 Super-resolution via semidenite programming High res Signal Estimate 3. Estimate amplitudes via least squares

91 Sparsity is not enough Theory Proof (sketch) Implementation via semidenite programming Robustness to noise

92 Estimation from noisy data Without noise, we achieve perfect precision, i.e. innite resolution y = F c x

93 Estimation from noisy data Without noise, we achieve perfect precision, i.e. innite resolution In practice, there is always noise y = F c x y = F c x + z Understanding the performance in a noisy setting is crucial for applications

94 Estimation from noisy data Without noise, we achieve perfect precision, i.e. innite resolution In practice, there is always noise y = F c x y = F c x + z Understanding the performance in a noisy setting is crucial for applications Metrics : 1. Approximation error at a higher resolution 2. Support-detection error

95 Super-resolution factor : spatial viewpoint Super-resolution factor SRF = λ c λ f

96 Super-resolution factor : spectral viewpoint Super-resolution factor SRF = f f c

97 Approximation at a higher resolution Resolution at scale λ is quantied by convolution with kernel φ λ of width λ At the resolution of the measurements φ λc (x est x) δ L 1

98 Approximation at a higher resolution Resolution at scale λ is quantied by convolution with kernel φ λ of width λ At the resolution of the measurements φ λc (x est x) δ L 1 How does the estimate degrade at a higher resolution?

99 Approximation at a higher resolution Resolution at scale λ is quantied by convolution with kernel φ λ of width λ At the resolution of the measurements φ λc (x est x) δ L 1 How does the estimate degrade at a higher resolution? Theorem [Candès, F. 2012] If 2 /f c then the solution ˆx to min x x TV subject to F c x y 2 δ, satises φλf (ˆx x) L1 SRF2 δ

100 Example SNR : 25 db Measurements High res signal

101 Example SNR : 25 db High res Signal Estimate

102 Example SNR : 25 db Noisy measurements Low res signal High res signal Estimate

103 Support detection Original signal, support T x = j a j δ tj a j C, t j T [0, 1]

104 Support detection Original signal, support T x = j a j δ tj a j C, t j T [0, 1] Estimated signal, support T ˆx = j â j δˆtj â j C, ˆtj T [0, 1]

105 Support detection Original signal, support T x = j a j δ tj a j C, t j T [0, 1] Estimated signal, support T ˆx = j â j δˆtj â j C, ˆtj T [0, 1] How accurately can we detect the support at a certain noise level δ?

106 Support-detection accuracy Theorem [F. 2013] For any t i T, if a i > C 1 δ there exists ˆti T such that ti ˆti 1 C 2 δ f c a i C 1 δ

107 Support-detection accuracy Theorem [F. 2013] For any t i T, if a i > C 1 δ there exists ˆti T such that ti ˆti 1 C 2 δ f c a i C 1 δ The support-detection accuracy is not aected by aliasing (no dependence on the amplitude of the signal at other locations)

108 Consequence Robustness of the algorithm to high dynamic ranges Measurements High res signal

109 Consequence Robustness of the algorithm to high dynamic ranges High res Signal Estimate (Real)

110 Conclusion and future work To obtain theoretical guarantees for super-resolution in realistic settings, we need conditions that avoid clustered supports

111 Conclusion and future work To obtain theoretical guarantees for super-resolution in realistic settings, we need conditions that avoid clustered supports Under a minimum-separation condition, convex programming achieves exact recovery

112 Conclusion and future work To obtain theoretical guarantees for super-resolution in realistic settings, we need conditions that avoid clustered supports Under a minimum-separation condition, convex programming achieves exact recovery The optimization problem can be recast as a tractable semidenite program

113 Conclusion and future work To obtain theoretical guarantees for super-resolution in realistic settings, we need conditions that avoid clustered supports Under a minimum-separation condition, convex programming achieves exact recovery The optimization problem can be recast as a tractable semidenite program The method is provably robust to noise

114 Conclusion and future work To obtain theoretical guarantees for super-resolution in realistic settings, we need conditions that avoid clustered supports Under a minimum-separation condition, convex programming achieves exact recovery The optimization problem can be recast as a tractable semidenite program The method is provably robust to noise Research directions : Super-resolution of images with sharp edges Developing fast solvers to solve sdp formulation Extending results to other overcomplete dictionaries

115 For more details Towards a mathematical theory of super-resolution. E. J. Candès and C. Fernandez-Granda. Comm. on Pure and Applied Math. Super-resolution from noisy data. E. J. Candès and C. Fernandez-Granda. Journal of Fourier Analysis and Applications Support detection in super-resolution. C. Fernandez-Granda. Proceedings of SampTA 2013 Prolate spheroidal wave functions, Fourier analysis, and uncertainty V - The discrete case. D. Slepian. Bell System Technical Journal, 57 : , 1978

Super-resolution via Convex Programming

Super-resolution via Convex Programming Super-resolution via Convex Programming Carlos Fernandez-Granda (Joint work with Emmanuel Candès) Structure and Randomness in System Identication and Learning, IPAM 1/17/2013 1/17/2013 1 / 44 Index 1 Motivation

More information

Support Detection in Super-resolution

Support Detection in Super-resolution Support Detection in Super-resolution Carlos Fernandez-Granda (Stanford University) 7/2/2013 SampTA 2013 Support Detection in Super-resolution C. Fernandez-Granda 1 / 25 Acknowledgements This work was

More information

Optimization-based sparse recovery: Compressed sensing vs. super-resolution

Optimization-based sparse recovery: Compressed sensing vs. super-resolution Optimization-based sparse recovery: Compressed sensing vs. super-resolution Carlos Fernandez-Granda, Google Computational Photography and Intelligent Cameras, IPAM 2/5/2014 This work was supported by a

More information

Sparse Recovery Beyond Compressed Sensing

Sparse Recovery Beyond Compressed Sensing Sparse Recovery Beyond Compressed Sensing Carlos Fernandez-Granda www.cims.nyu.edu/~cfgranda Applied Math Colloquium, MIT 4/30/2018 Acknowledgements Project funded by NSF award DMS-1616340 Separable Nonlinear

More information

Super-Resolution of Point Sources via Convex Programming

Super-Resolution of Point Sources via Convex Programming Super-Resolution of Point Sources via Convex Programming Carlos Fernandez-Granda July 205; Revised December 205 Abstract We consider the problem of recovering a signal consisting of a superposition of

More information

Super-resolution of point sources via convex programming

Super-resolution of point sources via convex programming Information and Inference: A Journal of the IMA 06) 5, 5 303 doi:0.093/imaiai/iaw005 Advance Access publication on 0 April 06 Super-resolution of point sources via convex programming Carlos Fernandez-Granda

More information

Demixing Sines and Spikes: Robust Spectral Super-resolution in the Presence of Outliers

Demixing Sines and Spikes: Robust Spectral Super-resolution in the Presence of Outliers Demixing Sines and Spikes: Robust Spectral Super-resolution in the Presence of Outliers Carlos Fernandez-Granda, Gongguo Tang, Xiaodong Wang and Le Zheng September 6 Abstract We consider the problem of

More information

Combining Sparsity with Physically-Meaningful Constraints in Sparse Parameter Estimation

Combining Sparsity with Physically-Meaningful Constraints in Sparse Parameter Estimation UIUC CSL Mar. 24 Combining Sparsity with Physically-Meaningful Constraints in Sparse Parameter Estimation Yuejie Chi Department of ECE and BMI Ohio State University Joint work with Yuxin Chen (Stanford).

More information

Going off the grid. Benjamin Recht Department of Computer Sciences University of Wisconsin-Madison

Going off the grid. Benjamin Recht Department of Computer Sciences University of Wisconsin-Madison Going off the grid Benjamin Recht Department of Computer Sciences University of Wisconsin-Madison Joint work with Badri Bhaskar Parikshit Shah Gonnguo Tang We live in a continuous world... But we work

More information

ROBUST BLIND SPIKES DECONVOLUTION. Yuejie Chi. Department of ECE and Department of BMI The Ohio State University, Columbus, Ohio 43210

ROBUST BLIND SPIKES DECONVOLUTION. Yuejie Chi. Department of ECE and Department of BMI The Ohio State University, Columbus, Ohio 43210 ROBUST BLIND SPIKES DECONVOLUTION Yuejie Chi Department of ECE and Department of BMI The Ohio State University, Columbus, Ohio 4 ABSTRACT Blind spikes deconvolution, or blind super-resolution, deals with

More information

On the recovery of measures without separation conditions

On the recovery of measures without separation conditions Norbert Wiener Center Department of Mathematics University of Maryland, College Park http://www.norbertwiener.umd.edu Applied and Computational Mathematics Seminar Georgia Institute of Technology October

More information

Sparse linear models

Sparse linear models Sparse linear models Optimization-Based Data Analysis http://www.cims.nyu.edu/~cfgranda/pages/obda_spring16 Carlos Fernandez-Granda 2/22/2016 Introduction Linear transforms Frequency representation Short-time

More information

Constrained optimization

Constrained optimization Constrained optimization DS-GA 1013 / MATH-GA 2824 Optimization-based Data Analysis http://www.cims.nyu.edu/~cfgranda/pages/obda_fall17/index.html Carlos Fernandez-Granda Compressed sensing Convex constrained

More information

A CONVEX-PROGRAMMING FRAMEWORK FOR SUPER-RESOLUTION

A CONVEX-PROGRAMMING FRAMEWORK FOR SUPER-RESOLUTION A CONVEX-PROGRAMMING FRAMEWORK FOR SUPER-RESOLUTION A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT

More information

Overview. Optimization-Based Data Analysis. Carlos Fernandez-Granda

Overview. Optimization-Based Data Analysis.   Carlos Fernandez-Granda Overview Optimization-Based Data Analysis http://www.cims.nyu.edu/~cfgranda/pages/obda_spring16 Carlos Fernandez-Granda 1/25/2016 Sparsity Denoising Regression Inverse problems Low-rank models Matrix completion

More information

Lecture Notes 9: Constrained Optimization

Lecture Notes 9: Constrained Optimization Optimization-based data analysis Fall 017 Lecture Notes 9: Constrained Optimization 1 Compressed sensing 1.1 Underdetermined linear inverse problems Linear inverse problems model measurements of the form

More information

Super-resolution by means of Beurling minimal extrapolation

Super-resolution by means of Beurling minimal extrapolation Super-resolution by means of Beurling minimal extrapolation Norbert Wiener Center Department of Mathematics University of Maryland, College Park http://www.norbertwiener.umd.edu Acknowledgements ARO W911NF-16-1-0008

More information

Going off the grid. Benjamin Recht University of California, Berkeley. Joint work with Badri Bhaskar Parikshit Shah Gonnguo Tang

Going off the grid. Benjamin Recht University of California, Berkeley. Joint work with Badri Bhaskar Parikshit Shah Gonnguo Tang Going off the grid Benjamin Recht University of California, Berkeley Joint work with Badri Bhaskar Parikshit Shah Gonnguo Tang imaging astronomy seismology spectroscopy x(t) = kx j= c j e i2 f jt DOA Estimation

More information

Strengthened Sobolev inequalities for a random subspace of functions

Strengthened Sobolev inequalities for a random subspace of functions Strengthened Sobolev inequalities for a random subspace of functions Rachel Ward University of Texas at Austin April 2013 2 Discrete Sobolev inequalities Proposition (Sobolev inequality for discrete images)

More information

Three Recent Examples of The Effectiveness of Convex Programming in the Information Sciences

Three Recent Examples of The Effectiveness of Convex Programming in the Information Sciences Three Recent Examples of The Effectiveness of Convex Programming in the Information Sciences Emmanuel Candès International Symposium on Information Theory, Istanbul, July 2013 Three stories about a theme

More information

ELE 538B: Sparsity, Structure and Inference. Super-Resolution. Yuxin Chen Princeton University, Spring 2017

ELE 538B: Sparsity, Structure and Inference. Super-Resolution. Yuxin Chen Princeton University, Spring 2017 ELE 538B: Sparsity, Structure and Inference Super-Resolution Yuxin Chen Princeton University, Spring 2017 Outline Classical methods for parameter estimation Polynomial method: Prony s method Subspace method:

More information

Robust Uncertainty Principles: Exact Signal Reconstruction from Highly Incomplete Frequency Information

Robust Uncertainty Principles: Exact Signal Reconstruction from Highly Incomplete Frequency Information 1 Robust Uncertainty Principles: Exact Signal Reconstruction from Highly Incomplete Frequency Information Emmanuel Candès, California Institute of Technology International Conference on Computational Harmonic

More information

Rapid, Robust, and Reliable Blind Deconvolution via Nonconvex Optimization

Rapid, Robust, and Reliable Blind Deconvolution via Nonconvex Optimization Rapid, Robust, and Reliable Blind Deconvolution via Nonconvex Optimization Shuyang Ling Department of Mathematics, UC Davis Oct.18th, 2016 Shuyang Ling (UC Davis) 16w5136, Oaxaca, Mexico Oct.18th, 2016

More information

Sparse Parameter Estimation: Compressed Sensing meets Matrix Pencil

Sparse Parameter Estimation: Compressed Sensing meets Matrix Pencil Sparse Parameter Estimation: Compressed Sensing meets Matrix Pencil Yuejie Chi Departments of ECE and BMI The Ohio State University Colorado School of Mines December 9, 24 Page Acknowledgement Joint work

More information

Recovering overcomplete sparse representations from structured sensing

Recovering overcomplete sparse representations from structured sensing Recovering overcomplete sparse representations from structured sensing Deanna Needell Claremont McKenna College Feb. 2015 Support: Alfred P. Sloan Foundation and NSF CAREER #1348721. Joint work with Felix

More information

Agenda. Applications of semidefinite programming. 1 Control and system theory. 2 Combinatorial and nonconvex optimization

Agenda. Applications of semidefinite programming. 1 Control and system theory. 2 Combinatorial and nonconvex optimization Agenda Applications of semidefinite programming 1 Control and system theory 2 Combinatorial and nonconvex optimization 3 Spectral estimation & super-resolution Control and system theory SDP in wide use

More information

AN ALGORITHM FOR EXACT SUPER-RESOLUTION AND PHASE RETRIEVAL

AN ALGORITHM FOR EXACT SUPER-RESOLUTION AND PHASE RETRIEVAL 2014 IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP) AN ALGORITHM FOR EXACT SUPER-RESOLUTION AND PHASE RETRIEVAL Yuxin Chen Yonina C. Eldar Andrea J. Goldsmith Department

More information

New Coherence and RIP Analysis for Weak. Orthogonal Matching Pursuit

New Coherence and RIP Analysis for Weak. Orthogonal Matching Pursuit New Coherence and RIP Analysis for Wea 1 Orthogonal Matching Pursuit Mingrui Yang, Member, IEEE, and Fran de Hoog arxiv:1405.3354v1 [cs.it] 14 May 2014 Abstract In this paper we define a new coherence

More information

Super-Resolution from Noisy Data

Super-Resolution from Noisy Data Super-Resolution from Noisy Data Emmanuel J. Candès and Carlos Fernandez-Granda October 2012; Revised April 2013 Abstract his paper studies the recovery of a superposition of point sources from noisy bandlimited

More information

The uniform uncertainty principle and compressed sensing Harmonic analysis and related topics, Seville December 5, 2008

The uniform uncertainty principle and compressed sensing Harmonic analysis and related topics, Seville December 5, 2008 The uniform uncertainty principle and compressed sensing Harmonic analysis and related topics, Seville December 5, 2008 Emmanuel Candés (Caltech), Terence Tao (UCLA) 1 Uncertainty principles A basic principle

More information

Random projections. 1 Introduction. 2 Dimensionality reduction. Lecture notes 5 February 29, 2016

Random projections. 1 Introduction. 2 Dimensionality reduction. Lecture notes 5 February 29, 2016 Lecture notes 5 February 9, 016 1 Introduction Random projections Random projections are a useful tool in the analysis and processing of high-dimensional data. We will analyze two applications that use

More information

INDUSTRIAL MATHEMATICS INSTITUTE. B.S. Kashin and V.N. Temlyakov. IMI Preprint Series. Department of Mathematics University of South Carolina

INDUSTRIAL MATHEMATICS INSTITUTE. B.S. Kashin and V.N. Temlyakov. IMI Preprint Series. Department of Mathematics University of South Carolina INDUSTRIAL MATHEMATICS INSTITUTE 2007:08 A remark on compressed sensing B.S. Kashin and V.N. Temlyakov IMI Preprint Series Department of Mathematics University of South Carolina A remark on compressed

More information

Part IV Compressed Sensing

Part IV Compressed Sensing Aisenstadt Chair Course CRM September 2009 Part IV Compressed Sensing Stéphane Mallat Centre de Mathématiques Appliquées Ecole Polytechnique Conclusion to Super-Resolution Sparse super-resolution is sometime

More information

Compressed sensing for radio interferometry: spread spectrum imaging techniques

Compressed sensing for radio interferometry: spread spectrum imaging techniques Compressed sensing for radio interferometry: spread spectrum imaging techniques Y. Wiaux a,b, G. Puy a, Y. Boursier a and P. Vandergheynst a a Institute of Electrical Engineering, Ecole Polytechnique Fédérale

More information

Compressed Sensing Off the Grid

Compressed Sensing Off the Grid Compressed Sensing Off the Grid Gongguo Tang, Badri Narayan Bhaskar, Parikshit Shah, and Benjamin Recht Department of Electrical and Computer Engineering Department of Computer Sciences University of Wisconsin-adison

More information

A Super-Resolution Algorithm for Multiband Signal Identification

A Super-Resolution Algorithm for Multiband Signal Identification A Super-Resolution Algorithm for Multiband Signal Identification Zhihui Zhu, Dehui Yang, Michael B. Wakin, and Gongguo Tang Department of Electrical Engineering, Colorado School of Mines {zzhu, dyang,

More information

Generalized Line Spectral Estimation via Convex Optimization

Generalized Line Spectral Estimation via Convex Optimization Generalized Line Spectral Estimation via Convex Optimization Reinhard Heckel and Mahdi Soltanolkotabi September 26, 206 Abstract Line spectral estimation is the problem of recovering the frequencies and

More information

arxiv: v3 [math.oc] 15 Sep 2014

arxiv: v3 [math.oc] 15 Sep 2014 Exact Support Recovery for Sparse Spikes Deconvolution arxiv:136.699v3 [math.oc] 15 Sep 214 Vincent Duval and Gabriel Peyré CNRS and Université Paris-Dauphine {vincent.duval,gabriel.peyre}@ceremade.dauphine.fr

More information

Near Optimal Signal Recovery from Random Projections

Near Optimal Signal Recovery from Random Projections 1 Near Optimal Signal Recovery from Random Projections Emmanuel Candès, California Institute of Technology Multiscale Geometric Analysis in High Dimensions: Workshop # 2 IPAM, UCLA, October 2004 Collaborators:

More information

Gauge optimization and duality

Gauge optimization and duality 1 / 54 Gauge optimization and duality Junfeng Yang Department of Mathematics Nanjing University Joint with Shiqian Ma, CUHK September, 2015 2 / 54 Outline Introduction Duality Lagrange duality Fenchel

More information

Introduction How it works Theory behind Compressed Sensing. Compressed Sensing. Huichao Xue. CS3750 Fall 2011

Introduction How it works Theory behind Compressed Sensing. Compressed Sensing. Huichao Xue. CS3750 Fall 2011 Compressed Sensing Huichao Xue CS3750 Fall 2011 Table of Contents Introduction From News Reports Abstract Definition How it works A review of L 1 norm The Algorithm Backgrounds for underdetermined linear

More information

Noisy Signal Recovery via Iterative Reweighted L1-Minimization

Noisy Signal Recovery via Iterative Reweighted L1-Minimization Noisy Signal Recovery via Iterative Reweighted L1-Minimization Deanna Needell UC Davis / Stanford University Asilomar SSC, November 2009 Problem Background Setup 1 Suppose x is an unknown signal in R d.

More information

Minimizing the Difference of L 1 and L 2 Norms with Applications

Minimizing the Difference of L 1 and L 2 Norms with Applications 1/36 Minimizing the Difference of L 1 and L 2 Norms with Department of Mathematical Sciences University of Texas Dallas May 31, 2017 Partially supported by NSF DMS 1522786 2/36 Outline 1 A nonconvex approach:

More information

Parameter Estimation for Mixture Models via Convex Optimization

Parameter Estimation for Mixture Models via Convex Optimization Parameter Estimation for Mixture Models via Convex Optimization Yuanxin Li Department of Electrical and Computer Engineering The Ohio State University Columbus Ohio 432 Email: li.3822@osu.edu Yuejie Chi

More information

Introduction to Compressed Sensing

Introduction to Compressed Sensing Introduction to Compressed Sensing Alejandro Parada, Gonzalo Arce University of Delaware August 25, 2016 Motivation: Classical Sampling 1 Motivation: Classical Sampling Issues Some applications Radar Spectral

More information

arxiv:submit/ [cs.it] 25 Jul 2012

arxiv:submit/ [cs.it] 25 Jul 2012 Compressed Sensing off the Grid Gongguo Tang, Badri Narayan Bhaskar, Parikshit Shah, and Benjamin Recht University of Wisconsin-Madison July 25, 202 arxiv:submit/05227 [cs.it] 25 Jul 202 Abstract We consider

More information

Elec4621 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis

Elec4621 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis Elec461 Advanced Digital Signal Processing Chapter 11: Time-Frequency Analysis Dr. D. S. Taubman May 3, 011 In this last chapter of your notes, we are interested in the problem of nding the instantaneous

More information

Lecture Notes 5: Multiresolution Analysis

Lecture Notes 5: Multiresolution Analysis Optimization-based data analysis Fall 2017 Lecture Notes 5: Multiresolution Analysis 1 Frames A frame is a generalization of an orthonormal basis. The inner products between the vectors in a frame and

More information

Self-Calibration and Biconvex Compressive Sensing

Self-Calibration and Biconvex Compressive Sensing Self-Calibration and Biconvex Compressive Sensing Shuyang Ling Department of Mathematics, UC Davis July 12, 2017 Shuyang Ling (UC Davis) SIAM Annual Meeting, 2017, Pittsburgh July 12, 2017 1 / 22 Acknowledgements

More information

MIT 9.520/6.860, Fall 2017 Statistical Learning Theory and Applications. Class 19: Data Representation by Design

MIT 9.520/6.860, Fall 2017 Statistical Learning Theory and Applications. Class 19: Data Representation by Design MIT 9.520/6.860, Fall 2017 Statistical Learning Theory and Applications Class 19: Data Representation by Design What is data representation? Let X be a data-space X M (M) F (M) X A data representation

More information

IMAGE RECONSTRUCTION FROM UNDERSAMPLED FOURIER DATA USING THE POLYNOMIAL ANNIHILATION TRANSFORM

IMAGE RECONSTRUCTION FROM UNDERSAMPLED FOURIER DATA USING THE POLYNOMIAL ANNIHILATION TRANSFORM IMAGE RECONSTRUCTION FROM UNDERSAMPLED FOURIER DATA USING THE POLYNOMIAL ANNIHILATION TRANSFORM Anne Gelb, Rodrigo B. Platte, Rosie Renaut Development and Analysis of Non-Classical Numerical Approximation

More information

SUPER-RESOLUTION OF POSITIVE SOURCES: THE DISCRETE SETUP. Veniamin I. Morgenshtern Emmanuel J. Candès

SUPER-RESOLUTION OF POSITIVE SOURCES: THE DISCRETE SETUP. Veniamin I. Morgenshtern Emmanuel J. Candès SUPER-RESOLUTIO OF POSITIVE SOURCES: THE DISCRETE SETUP By Veniamin I. Morgenshtern Emmanuel J. Candès Technical Report o. 205-0 May 205 Department of Statistics STAFORD UIVERSITY Stanford, California

More information

Stable Signal Recovery from Incomplete and Inaccurate Measurements

Stable Signal Recovery from Incomplete and Inaccurate Measurements Stable Signal Recovery from Incomplete and Inaccurate Measurements EMMANUEL J. CANDÈS California Institute of Technology JUSTIN K. ROMBERG California Institute of Technology AND TERENCE TAO University

More information

Super-Resolution of Mutually Interfering Signals

Super-Resolution of Mutually Interfering Signals Super-Resolution of utually Interfering Signals Yuanxin Li Department of Electrical and Computer Engineering The Ohio State University Columbus Ohio 430 Email: li.38@osu.edu Yuejie Chi Department of Electrical

More information

Compressed Sensing and Neural Networks

Compressed Sensing and Neural Networks and Jan Vybíral (Charles University & Czech Technical University Prague, Czech Republic) NOMAD Summer Berlin, September 25-29, 2017 1 / 31 Outline Lasso & Introduction Notation Training the network Applications

More information

Robust Recovery of Positive Stream of Pulses

Robust Recovery of Positive Stream of Pulses 1 Robust Recovery of Positive Stream of Pulses Tamir Bendory arxiv:1503.0878v3 [cs.it] 9 Jan 017 Abstract The problem of estimating the delays and amplitudes of a positive stream of pulses appears in many

More information

Near Ideal Behavior of a Modified Elastic Net Algorithm in Compressed Sensing

Near Ideal Behavior of a Modified Elastic Net Algorithm in Compressed Sensing Near Ideal Behavior of a Modified Elastic Net Algorithm in Compressed Sensing M. Vidyasagar Cecil & Ida Green Chair The University of Texas at Dallas M.Vidyasagar@utdallas.edu www.utdallas.edu/ m.vidyasagar

More information

Exact Joint Sparse Frequency Recovery via Optimization Methods

Exact Joint Sparse Frequency Recovery via Optimization Methods IEEE TRANSACTIONS ON SIGNAL PROCESSING Exact Joint Sparse Frequency Recovery via Optimization Methods Zai Yang, Member, IEEE, and Lihua Xie, Fellow, IEEE arxiv:45.6585v [cs.it] 3 May 6 Abstract Frequency

More information

Approximate Support Recovery of Atomic Line Spectral Estimation: A Tale of Resolution and Precision

Approximate Support Recovery of Atomic Line Spectral Estimation: A Tale of Resolution and Precision Approximate Support Recovery of Atomic Line Spectral Estimation: A Tale of Resolution and Precision Qiuwei Li and Gongguo Tang Department of Electrical Engineering and Computer Science, Colorado School

More information

Information and Resolution

Information and Resolution Information and Resolution (Invited Paper) Albert Fannjiang Department of Mathematics UC Davis, CA 95616-8633. fannjiang@math.ucdavis.edu. Abstract The issue of resolution with complex-field measurement

More information

Primal Dual Pursuit A Homotopy based Algorithm for the Dantzig Selector

Primal Dual Pursuit A Homotopy based Algorithm for the Dantzig Selector Primal Dual Pursuit A Homotopy based Algorithm for the Dantzig Selector Muhammad Salman Asif Thesis Committee: Justin Romberg (Advisor), James McClellan, Russell Mersereau School of Electrical and Computer

More information

CS 229r: Algorithms for Big Data Fall Lecture 19 Nov 5

CS 229r: Algorithms for Big Data Fall Lecture 19 Nov 5 CS 229r: Algorithms for Big Data Fall 215 Prof. Jelani Nelson Lecture 19 Nov 5 Scribe: Abdul Wasay 1 Overview In the last lecture, we started discussing the problem of compressed sensing where we are given

More information

COMPRESSIVE SENSING WITH AN OVERCOMPLETE DICTIONARY FOR HIGH-RESOLUTION DFT ANALYSIS. Guglielmo Frigo and Claudio Narduzzi

COMPRESSIVE SENSING WITH AN OVERCOMPLETE DICTIONARY FOR HIGH-RESOLUTION DFT ANALYSIS. Guglielmo Frigo and Claudio Narduzzi COMPRESSIVE SESIG WITH A OVERCOMPLETE DICTIOARY FOR HIGH-RESOLUTIO DFT AALYSIS Guglielmo Frigo and Claudio arduzzi University of Padua Department of Information Engineering DEI via G. Gradenigo 6/b, I

More information

Compressive Line Spectrum Estimation with Clustering and Interpolation

Compressive Line Spectrum Estimation with Clustering and Interpolation Compressive Line Spectrum Estimation with Clustering and Interpolation Dian Mo Department of Electrical and Computer Engineering University of Massachusetts Amherst, MA, 01003 mo@umass.edu Marco F. Duarte

More information

New Applications of Sparse Methods in Physics. Ra Inta, Centre for Gravitational Physics, The Australian National University

New Applications of Sparse Methods in Physics. Ra Inta, Centre for Gravitational Physics, The Australian National University New Applications of Sparse Methods in Physics Ra Inta, Centre for Gravitational Physics, The Australian National University 2 Sparse methods A vector is S-sparse if it has at most S non-zero coefficients.

More information

Convex Optimization and l 1 -minimization

Convex Optimization and l 1 -minimization Convex Optimization and l 1 -minimization Sangwoon Yun Computational Sciences Korea Institute for Advanced Study December 11, 2009 2009 NIMS Thematic Winter School Outline I. Convex Optimization II. l

More information

Sparse DOA estimation with polynomial rooting

Sparse DOA estimation with polynomial rooting Sparse DOA estimation with polynomial rooting Angeliki Xenaki Department of Applied Mathematics and Computer Science Technical University of Denmark 28 Kgs. Lyngby, Denmark Email: anxe@dtu.dk Peter Gerstoft

More information

An iterative hard thresholding estimator for low rank matrix recovery

An iterative hard thresholding estimator for low rank matrix recovery An iterative hard thresholding estimator for low rank matrix recovery Alexandra Carpentier - based on a joint work with Arlene K.Y. Kim Statistical Laboratory, Department of Pure Mathematics and Mathematical

More information

Multichannel Sampling of Signals with Finite. Rate of Innovation

Multichannel Sampling of Signals with Finite. Rate of Innovation Multichannel Sampling of Signals with Finite 1 Rate of Innovation Hojjat Akhondi Asl, Pier Luigi Dragotti and Loic Baboulaz EDICS: DSP-TFSR, DSP-SAMP. Abstract In this paper we present a possible extension

More information

Solving Corrupted Quadratic Equations, Provably

Solving Corrupted Quadratic Equations, Provably Solving Corrupted Quadratic Equations, Provably Yuejie Chi London Workshop on Sparse Signal Processing September 206 Acknowledgement Joint work with Yuanxin Li (OSU), Huishuai Zhuang (Syracuse) and Yingbin

More information

STAT 200C: High-dimensional Statistics

STAT 200C: High-dimensional Statistics STAT 200C: High-dimensional Statistics Arash A. Amini May 30, 2018 1 / 57 Table of Contents 1 Sparse linear models Basis Pursuit and restricted null space property Sufficient conditions for RNS 2 / 57

More information

GREEDY SIGNAL RECOVERY REVIEW

GREEDY SIGNAL RECOVERY REVIEW GREEDY SIGNAL RECOVERY REVIEW DEANNA NEEDELL, JOEL A. TROPP, ROMAN VERSHYNIN Abstract. The two major approaches to sparse recovery are L 1-minimization and greedy methods. Recently, Needell and Vershynin

More information

PHASE RETRIEVAL OF SPARSE SIGNALS FROM MAGNITUDE INFORMATION. A Thesis MELTEM APAYDIN

PHASE RETRIEVAL OF SPARSE SIGNALS FROM MAGNITUDE INFORMATION. A Thesis MELTEM APAYDIN PHASE RETRIEVAL OF SPARSE SIGNALS FROM MAGNITUDE INFORMATION A Thesis by MELTEM APAYDIN Submitted to the Office of Graduate and Professional Studies of Texas A&M University in partial fulfillment of the

More information

Simulations of the Microbunching Instability in FEL Beam Delivery Systems

Simulations of the Microbunching Instability in FEL Beam Delivery Systems Simulations of the Microbunching Instability in FEL Beam Delivery Systems Ilya Pogorelov Tech-X Corporation Workshop on High Average Power & High Brightness Beams UCLA, January 2009 Outline The setting:

More information

Sparsity in system identification and data-driven control

Sparsity in system identification and data-driven control 1 / 40 Sparsity in system identification and data-driven control Ivan Markovsky This signal is not sparse in the "time domain" 2 / 40 But it is sparse in the "frequency domain" (it is weighted sum of six

More information

Solution-recovery in l 1 -norm for non-square linear systems: deterministic conditions and open questions

Solution-recovery in l 1 -norm for non-square linear systems: deterministic conditions and open questions Solution-recovery in l 1 -norm for non-square linear systems: deterministic conditions and open questions Yin Zhang Technical Report TR05-06 Department of Computational and Applied Mathematics Rice University,

More information

Signal Recovery, Uncertainty Relations, and Minkowski Dimension

Signal Recovery, Uncertainty Relations, and Minkowski Dimension Signal Recovery, Uncertainty Relations, and Minkowski Dimension Helmut Bőlcskei ETH Zurich December 2013 Joint work with C. Aubel, P. Kuppinger, G. Pope, E. Riegler, D. Stotz, and C. Studer Aim of this

More information

ECE 8201: Low-dimensional Signal Models for High-dimensional Data Analysis

ECE 8201: Low-dimensional Signal Models for High-dimensional Data Analysis ECE 8201: Low-dimensional Signal Models for High-dimensional Data Analysis Lecture 3: Sparse signal recovery: A RIPless analysis of l 1 minimization Yuejie Chi The Ohio State University Page 1 Outline

More information

Compressed sensing. Or: the equation Ax = b, revisited. Terence Tao. Mahler Lecture Series. University of California, Los Angeles

Compressed sensing. Or: the equation Ax = b, revisited. Terence Tao. Mahler Lecture Series. University of California, Los Angeles Or: the equation Ax = b, revisited University of California, Los Angeles Mahler Lecture Series Acquiring signals Many types of real-world signals (e.g. sound, images, video) can be viewed as an n-dimensional

More information

Multiresolution Analysis

Multiresolution Analysis Multiresolution Analysis DS-GA 1013 / MATH-GA 2824 Optimization-based Data Analysis http://www.cims.nyu.edu/~cfgranda/pages/obda_fall17/index.html Carlos Fernandez-Granda Frames Short-time Fourier transform

More information

Conditions for Robust Principal Component Analysis

Conditions for Robust Principal Component Analysis Rose-Hulman Undergraduate Mathematics Journal Volume 12 Issue 2 Article 9 Conditions for Robust Principal Component Analysis Michael Hornstein Stanford University, mdhornstein@gmail.com Follow this and

More information

arxiv: v2 [cs.it] 16 Feb 2013

arxiv: v2 [cs.it] 16 Feb 2013 Atomic norm denoising with applications to line spectral estimation Badri Narayan Bhaskar, Gongguo Tang, and Benjamin Recht Department of Electrical and Computer Engineering Department of Computer Sciences

More information

Lecture: Introduction to Compressed Sensing Sparse Recovery Guarantees

Lecture: Introduction to Compressed Sensing Sparse Recovery Guarantees Lecture: Introduction to Compressed Sensing Sparse Recovery Guarantees http://bicmr.pku.edu.cn/~wenzw/bigdata2018.html Acknowledgement: this slides is based on Prof. Emmanuel Candes and Prof. Wotao Yin

More information

Robust linear optimization under general norms

Robust linear optimization under general norms Operations Research Letters 3 (004) 50 56 Operations Research Letters www.elsevier.com/locate/dsw Robust linear optimization under general norms Dimitris Bertsimas a; ;, Dessislava Pachamanova b, Melvyn

More information

Signal Recovery from Permuted Observations

Signal Recovery from Permuted Observations EE381V Course Project Signal Recovery from Permuted Observations 1 Problem Shanshan Wu (sw33323) May 8th, 2015 We start with the following problem: let s R n be an unknown n-dimensional real-valued signal,

More information

arxiv: v2 [math.na] 4 Jun 2017

arxiv: v2 [math.na] 4 Jun 2017 Sampling the Fourier transform along radial lines Charles Dossal Vincent Duval Clarice Poon December, 2016; Revised May, 2017 arxiv:1612.06752v2 [math.na] 4 Jun 2017 Abstract This article considers the

More information

Sparse Optimization Lecture: Dual Certificate in l 1 Minimization

Sparse Optimization Lecture: Dual Certificate in l 1 Minimization Sparse Optimization Lecture: Dual Certificate in l 1 Minimization Instructor: Wotao Yin July 2013 Note scriber: Zheng Sun Those who complete this lecture will know what is a dual certificate for l 1 minimization

More information

Compressed Sensing and Sparse Recovery

Compressed Sensing and Sparse Recovery ELE 538B: Sparsity, Structure and Inference Compressed Sensing and Sparse Recovery Yuxin Chen Princeton University, Spring 217 Outline Restricted isometry property (RIP) A RIPless theory Compressed sensing

More information

Sensing systems limited by constraints: physical size, time, cost, energy

Sensing systems limited by constraints: physical size, time, cost, energy Rebecca Willett Sensing systems limited by constraints: physical size, time, cost, energy Reduce the number of measurements needed for reconstruction Higher accuracy data subject to constraints Original

More information

The Pros and Cons of Compressive Sensing

The Pros and Cons of Compressive Sensing The Pros and Cons of Compressive Sensing Mark A. Davenport Stanford University Department of Statistics Compressive Sensing Replace samples with general linear measurements measurements sampled signal

More information

MATH 829: Introduction to Data Mining and Analysis Support vector machines and kernels

MATH 829: Introduction to Data Mining and Analysis Support vector machines and kernels 1/12 MATH 829: Introduction to Data Mining and Analysis Support vector machines and kernels Dominique Guillot Departments of Mathematical Sciences University of Delaware March 14, 2016 Separating sets:

More information

Two added structures in sparse recovery: nonnegativity and disjointedness. Simon Foucart University of Georgia

Two added structures in sparse recovery: nonnegativity and disjointedness. Simon Foucart University of Georgia Two added structures in sparse recovery: nonnegativity and disjointedness Simon Foucart University of Georgia Semester Program on High-Dimensional Approximation ICERM 7 October 2014 Part I: Nonnegative

More information

Fast and Robust Phase Retrieval

Fast and Robust Phase Retrieval Fast and Robust Phase Retrieval Aditya Viswanathan aditya@math.msu.edu CCAM Lunch Seminar Purdue University April 18 2014 0 / 27 Joint work with Yang Wang Mark Iwen Research supported in part by National

More information

COMPRESSED SENSING IN PYTHON

COMPRESSED SENSING IN PYTHON COMPRESSED SENSING IN PYTHON Sercan Yıldız syildiz@samsi.info February 27, 2017 OUTLINE A BRIEF INTRODUCTION TO COMPRESSED SENSING A BRIEF INTRODUCTION TO CVXOPT EXAMPLES A Brief Introduction to Compressed

More information

Boxlets: a Fast Convolution Algorithm for. Signal Processing and Neural Networks. Patrice Y. Simard, Leon Bottou, Patrick Haner and Yann LeCun

Boxlets: a Fast Convolution Algorithm for. Signal Processing and Neural Networks. Patrice Y. Simard, Leon Bottou, Patrick Haner and Yann LeCun Boxlets: a Fast Convolution Algorithm for Signal Processing and Neural Networks Patrice Y. Simard, Leon Bottou, Patrick Haner and Yann LeCun AT&T Labs-Research 100 Schultz Drive, Red Bank, NJ 07701-7033

More information

SPECTRAL COMPRESSIVE SENSING WITH POLAR INTERPOLATION. Karsten Fyhn, Hamid Dadkhahi, Marco F. Duarte

SPECTRAL COMPRESSIVE SENSING WITH POLAR INTERPOLATION. Karsten Fyhn, Hamid Dadkhahi, Marco F. Duarte SPECTRAL COMPRESSIVE SENSING WITH POLAR INTERPOLATION Karsten Fyhn, Hamid Dadkhahi, Marco F. Duarte Dept. of Electronic Systems, Aalborg University, Denmark. Dept. of Electrical and Computer Engineering,

More information

Optimized Compact-support Interpolation Kernels

Optimized Compact-support Interpolation Kernels Optimized Compact-support Interpolation Kernels Ramtin Madani, Student Member, IEEE, Ali Ayremlou, Student Member, IEEE, Arash Amini, Farrokh Marvasti, Senior Member, IEEE, Abstract In this paper, we investigate

More information

The convex algebraic geometry of rank minimization

The convex algebraic geometry of rank minimization The convex algebraic geometry of rank minimization Pablo A. Parrilo Laboratory for Information and Decision Systems Massachusetts Institute of Technology International Symposium on Mathematical Programming

More information

Gaussian Phase Transitions and Conic Intrinsic Volumes: Steining the Steiner formula

Gaussian Phase Transitions and Conic Intrinsic Volumes: Steining the Steiner formula Gaussian Phase Transitions and Conic Intrinsic Volumes: Steining the Steiner formula Larry Goldstein, University of Southern California Nourdin GIoVAnNi Peccati Luxembourg University University British

More information

Enhanced Compressive Sensing and More

Enhanced Compressive Sensing and More Enhanced Compressive Sensing and More Yin Zhang Department of Computational and Applied Mathematics Rice University, Houston, Texas, U.S.A. Nonlinear Approximation Techniques Using L1 Texas A & M University

More information