Reading Seminar on
Data Science and Compressive Sensing

formerly Compressive Sensing, Extensions, and Applications


Wednesdays at 5:30pm via Zoom.
  • 24 Mar: Chunyang Liao presented When do neural networks outperform kernel methods? by B. Ghorbani, S. Mei, T. Misiakiewicz, and A. Montanari. (arXiv)
  • 17 Mar: Ryan Malthaner presented On the stability properties and the optimization landscape of training problems with squared loss for neural networks and general nonlinear conic approximation schemes. by C. Christof (arXiv)
  • 03 Mar: Bolong Ma presented Spurious valleys in one-hidden-layer neural network optimization landscapes by L. Venturi, A. Bandeira, J. Bruna. (link)
  • 24 Feb: Mahmood Ettehad presented A well-tempered landscape for non-convex robust subspace recovery by T. Maunu, T. Zhang, G. Lerman. (link)
  • 10 Feb: Chunyang Liao presented Optimal estimation of linear operators in Hilbert spaces from inaccurate data by A. Melkman and C. Micchelli (doi) and Optimal estimation of linear operators from inaccurate data: A second look by C. Micchelli (doi).
  • 03 Feb: Tushar Pandey presented Explicit frames for deterministic phase retrieval via PhaseLift by M. Kech. (doi)
  • 27 Jan: Simon Foucart presented Optimal approximation of continuous functions by very deep ReLU networks by D. Yarotsky. (link)

Wednesdays at 5:00pm via Zoom.
  • 28 Oct: Kung-Ching Lin presented Nearly-tight VC-dimension and pseudodimension bounds for piecewise linear neural networks by P. Bartlett, N. Harvey, C. Liaw, and A. Mehrabian. (link)
  • 21 Oct: Josiah Park presented Polynomial codes: an optimal design for high-dimensional coded matrix multiplication by Q. Yu, M. Ali Maddah-Ali, and A. S. Avestimehr. (arXiv)
  • 14 Oct: Mahmood Ettehad presented Nonlinear reduced models for state and parameter estimation by A. Cohen, W. Dahmen, O.  Mula, and J. Nichols. (arXiv)
  • 07 Oct: Tushar Pandey presented Unbalanced expanders and randomness extractors from Parvesh-Vardy codes by V. Guruswami, C. Vadhan, and S. Umans. (doi)
  • 30 Sep: Bolong Ma presented Dense quantum measurement theory by L. Gyongyosi and S. Imre. (doi)
  • 23 Sep: Ryan Malthaner presented Memory capacity of neural networks with threshold and ReLU activations by R. Vershynin (arXiv)
  • 16 Sep: Srinivas Subramanian presented Efficient projections onto the l1-ball for learning in high dimensions by J. Duchi, S. Shalev-Shwartz, Y. Singer and T. Chandra (doi) and Sparse projections onto the simplex by A. Kyrillidis, S. Becker, V. Cevher, and C. Koch (link)
  • 09 Sep: Chunyang Liao presented Regularization in regression with bounded noise: a Chebyshev center approach by A. Beck and Y. Eldar. (doi)
  • 02 Sep: Kung-Ching Lin talked about Signal decimation with optimal reconstruction error guarantee.
  • 26 Aug: Simon Foucart presented Instances of computational optimal recovery: refined approximability models. (doi)

Tuesdays at 4:00pm in Blocker 608M.
  • 03 Mar: Chunyang Liao presented Learning to benchmark: determining best achievable misclassification error from training data by M. Noshad, L. Xu, and A. Hero. (arXiv)
  • 25 Feb: Bolong Ma presented Multilinear compressive sensing and an application to convolutional linear networks by F.Malgouyres and J. Landsberg. (doi)
  • 18 Feb: Mahmood Ettehad presented Optimal sampling rates for approximating analytic functions from pointwise samples by B. Adcock, R. Platte, and A. Shadrin. (doi)
  • 11 Feb: Srinivas Subramanian presented Multi-layer sparse coding: the holistic way by A. Aberdam, J. Sulam, and M. Elad. (doi)
  • 04 Feb: Ryan Malthaner presented Deep network approximation for smooth functions by J. Lu, Z. Shen, H. Yang, and S. Zhang. (arXiv)
  • 28 Jan: Simon Foucart presented Facilitating OWL norm minimizations.

Wednesdays at 5:00pm in Blocker 608M.
  • 20 Nov: Ming Wei presented Universal matrix completion, by S. Bhojanapalli and P. Jain. (arXiv)
  • 13 Nov: Qi Yuan presented Random gradient extrapolation for distributed and stochastic optimization, by G. Lan and Y. Zhou. (doi)
  • 06 Nov: Jiangyuan Li presented Surprises in High-Dimensional Ridgeless Least Squares Interpolation, by T. Hastie, A. Montanari, S. Rosset, and R. J. Tibshirani. (arXiv)
  • 30 Oct: Srinivas Subramanian presented Reliable recovery of hierarchically sparse signals and application in machine-type communications, by I. Roth, M. Kliesch, G. Wunder, and J. Eisert (arXiv)
  • 23 Oct: Simon Foucart discussed Determinism in compressive data acquisition.
  • 09 Oct: Chunyang Liao presented ReLU deep neural networks and linear finite elements, by J. He, L. Li, J. Xu, and C. Zheng. (arXiv)
  • 02 Oct: Ryan Malthaner presented Nonlinear approximation via compositions, by Z. Shen, H. Yang, and S. Zhang. (arXiv)
  • 25 Sep: Bolong Ma presented Deep neural network approximation theory, by P. Grohs, D. Perekrestenko, D. Elbrächter, and H. Bölcskei. (arXiv)
  • 18 Sep: Mahmood Ettehad presented Just interpolate: kernel "ridgeless" regression can generalize, by T. Liang and A. Rakhlin. (arXiv)
  • 11 Sep: Srinivas Subramanian presented Algorithmic regularization in over-parameterized matrix sensing and neural networks with quadratic activations, by Y. Li, T. Ma, and H. Zhang. (link)
  • 04 Sep: Simon Foucart presented Sampling schemes and recovery algorithms for functions of few coordinate variables.

Fridays at 3:40pm in Blocker 608M.
  • 02 Nov: Jiangyuan Li presented Algorithmic aspects of inverse problems using generative models, by C. Hegde. (arXiv)
  • 26 Oct: Rick Lynch presented Preserving injectivity under subgaussian mappings and its application to compressed sensing, by G. Casazza, X. Chen, and himself. (arXiv)
  • 19 Oct: Srinivas Subramanian presented Modeling sparse deviations for compressed sensing using generative models, by M. Dhar, A. Grover, and S. Ermon. (arXiv)
  • 05 Oct: Ryan Malthaner presented A compressed sensing view of unsupervised text embeddings, bag-of-n-grams, and LSTMs, by S. Arora, M. Khodak, N. Saunshi, and K. Vodrahalli. (link)
  • 28 Sep: Simon Foucart talked about Semidefinite programming in approximation theory: two examples.
  • 21 Sep: Bolong Ma presented Mad Max: affine spline insights into deep learning, by R. Balestriero and R. Baraniuk. (arXiv)
  • 14 Sep: Mahmood Ettehad presented Learning without mixing: towards a sharp analysis of linear system identification, by M. Simchowitz, H. Mania, S. Tu, M. Jordan, and B. Recht. (arXiv)

Thursdays at 4:00pm, usually in Blocker 608M.
  • 19 Apr: Ryan Malthaner presented Generalization in machine learning via analytical learning theory, by K. Kawaguchi and Y. Bengio. (arXiv)
  • 12 Apr: Bolong Ma presented Constructive methods of approximation by ridge functions and radial functions by Y. Xu, W. Light, and W. Cheney. (doi)
  • 05 Apr: Xiaojing Wang talked about Seismic data denoising based on dictionary learning. In Blocker 624.
  • 08 Mar: Laurent Jacques talked about Time for dithering! Quantized random embeddings with RIP random matrices. In Blocker 628.
  • 01 Mar: Xiaohan Chen talked about Deep learning techniques in compressive sensing and optimization.
  • 22 Feb: Rick Lynch presented Sparse recovery under weak moment assumptions, by G. Lecué and S. Mendelson. (doi)
  • 08 Feb: Srinivas Subramanian presented Normalized iterative hard thresholding for matrix completion, by J. Tanner and K. Wei. (doi)
  • 01 Feb: Mahmood Ettehad presented Data assimilation in reduced modeling, by P. Binev, A. Cohen, W. Dahmen, R. DeVore, G. Petrova, and P. Wojtaszczyk. (doi)
  • 25 Jan: Jiangyuan Li presented Matrix completion via max-norm constrained optimization, by T. Cai and W.-X. Zhou. (doi)

Fridays at 4:00pm in Blocker 608M.
  • 01 Dec: Mahmood Ettehad presented Linear system identification via atomic norm regularization, by P. Shah, B. N. Bhaskar, G. Tang and B. Recht. (doi)
  • 17 Nov: Bolong Ma presented Optimal approximation with sparsely connected deep neural networks, by H. Bölcskei, P. Grohs, G. Kutyniok, P. Petersen. (arXiv)
  • 10 Nov: Srinivas Subramanian presented Cubature, approximation, and isotropy in the hypercube, by Lloyd N. Trefethen. (doi)
  • 03 Nov: Rick Lynch presented Towards understanding the invertibility of convolutional neural networks, by A. Gilbert, Y. Zhang, K. Lee, Y. Zhang, H. Lee. (arXiv)
  • 27 Oct: Simon Foucart presented Infinite-dimensional L1 minimization and function approximation from pointwise data, by B. Adcock. (doi)
  • 13 Oct: Jiangyuan Li presented Stable optimizationless recovery from phaseless linear measurements, by L. Demanet and P. Hand. (doi)
  • 29 Sep: Bolong Ma presented A mathematical theory of deep convolutional neural networks for feature extraction, by T. Wiatowski, H. Bölcskei. (arXiv)
  • 22 Sep: Rick Lynch presented Compressed sensing using generative models, by A. Bora, A. Jalal, E. Price, A. Dimakis. (arXiv)
  • 15 Sep: Srinivas Subramanian presented Compressed sensing from phaseless gaussian measurements via linear programming in the natural parameter space, by P. Hand and V. Voroninski. (arXiv)
  • 08 Sep: Mahmood Ettehad presented Non-asymptotic analysis of robust control from coarse-grained identification, by S. Tu, R. Boczar, A. Packard, and B. Recht. (arXiv)
  • 01 Sep: Simon Foucart presented Solving random quadratic systems of equations is nearly as easy as solving linear systems, by Y. Chen and E. Candès. (doi)

Fridays at 3:00pm in Blocker 608M.
  • 21 Apr: Simon Foucart presented An IHT algorithm for sparse recovery from subexponential measurements, by S. Foucart and G. Lecué.
  • 03 Mar: Jiangyuan Li presented One-bit compressed sensing with non-Gaussian measurements, by A. Ai, A. Lapanowski, Y. Plan, and R. Vershynin. (doi)
  • 24 Feb: Mahmood Ettehad presented A simpler approach to matrix completion, by B. Recht. (arXiv)
  • 17 Feb: Srinivas Subramanian presented Recent developments in the sparse Fourier transform: a compressed Fourier transform for big data, by A. Gilbert, P. Indyk, M. Iwen, and L. Schmidt. (doi)
  • 10 Feb: Xinjie Fan presented Minimax lower bounds on dictionary learning for tensor data, by Z. Shakeri, W. Bajwa, and A. Sarwate. (arXiv)
  • 03 Feb: Rick Lynch presented Online dictionary learning for sparse coding, by J. Mairal, F. Bach, J. Ponce, and G. Sapiro. (doi)
  • 27 Jan: Simon Foucart presented Sparse recovery from saturated measurements, by S. Foucart and T. Needham. (doi)

Wednesdays at 4:00pm in Blocker 608M.
  • 07 Dec: Xinjie Fan presented The generalized Lasso with non-linear observations, by Y. Plan and R. Vershynin. (arXiv)
  • 30 Nov: Srinivas Subramanian presented Orthogonal matching pursuit under the restricted isometry property, by A. Cohen, W. Dahmen, and R. DeVore. (doi)
  • 16 Nov: Mahmood Ettehad presented Sparse and spurious: dictionary learning with noise and outliers, by R. Gribonval, R. Jenatton and F. Bach. (doi)
  • 9 Nov: Rick Lynch presented Compressive sensing with redundant dictionaries and structured measurements, by F. Krahmer, D. Needell, and R. Ward. (doi)
  • 2 Nov: Simon Foucart presented A least-squares method for sparse low rank approximation of multivariate functions, by M. Chevreuil, R. Lebrun, A. Nouy, and P. Rai. (arXiv)
  • 26 Oct: Xinjie Fan presented A Tight Bound of Hard Thresholding, by J. Shen and P. Li. (arXiv)
  • 19 Oct: Srinivas Subramanian presented GPU accelerated greedy algorithms for compressed sensing, by J. Blanchard and J. Tanner. (doi)
  • 12 Oct: Mahmood Ettehad presented Link delay estimation via expander graphs, by M. Firooz and S. Roy. (doi)
  • 5 Oct: Rick Lynch presented A conditional construction of restricted isometries, by A. Bandeira, D. Mixon, and J. Moreira. (arXiv)
  • 21 Sep: Simon Foucart presented Improving compressed sensing with the diamond norm, by M. Kliesch, R. Kueng, J. Eisert, and D. Gross. (arXiv)
  • 14 Sep: David Koslicki (Oregon State University) presented Quikr: a method for rapid reconstruction of bacterial communities via compressive sensing, by D. Koslicki, S. Foucart, and G. Rosen, as well as Sparse recovery by means of nonnegative least squares, by S. Foucart and D. Koslicki. (doi) (doi)

Thursdays at 3:00pm in Blocker 608M.
  • 21 Apr: Srinivas Subramanian presented Message-passing algorithms for compressed sensing, by D. L. Donoho, A. Maleki, and A. Montanari. (doi)
  • 14 Apr: Mahmood Ettehad presented Exact reconstruction of gene regulatory networks using compressive sensing, by Y. Chang, J. Gray, and C. Tomlin. (link)
  • 31 Mar: Xinjie Fan presented Robust subspace clustering, by M. Soltanolkotabi, E. Elhamifar, and E. Candès. (doi)
  • 24 Mar: Simon Foucart presented When are Quasi-Monte Carlo algorithms efficient for high dimensional integrals?, by I. Sloan and H. Woźniakowski. (doi)
  • 10 Mar: Srinivas Subramanian presented Expander l0-decoding, by R. Mendoza-Smith and J. Tanner. (arXiv)
  • 3 Mar: Mahmood Ettehad presented two papers by B. Sanandaji, T. Vincent, and M. Wakin. (doi, doi)
  • 25 Feb: Avinash Vem presented Sub-linear time compressed sensing for support recovery using sparse-graph codes by X. Li, S. Pawar, and K. Ramchandran. (arXiv)
  • 18 Feb: Xinjie Fan presented On the fundamental limits of adaptive sensing, by E. Arias-Castro, E. Candès, and M. Davenport. (arXiv)
  • 11 Feb: Simon Foucart presented Sparse disjointed recovery from noninflating measurements, by S. Foucart, M. Minner, and T. Needham. (doi)

Thursdays at 4:00pm in Blocker 628.
  • 01 Oct: Overview of the Mathematics of Compressive Sensing - Part 1, presented by Simon Foucart. (slides)
  • 08 Oct: Overview of the Mathematics of Compressive Sensing - Part 2, presented by Simon Foucart. (slides)
  • 15 Oct: Overview of the Mathematics of Compressive Sensing - Part 3, presented by Simon Foucart. (slides)
  • 29 Oct: Overview of the Mathematics of Compressive Sensing - Part 4, presented by Simon Foucart. (slides)
  • 05 Nov: Nathan LaFerney presented Living on the edge: phase transitions in convex programs with random data, by D. Amelunxen, M. Lotz, M. McCoy, and J. Tropp. (arXiv)
  • 19 Nov: Srinivas Subramanian presented Efficient and robust compressed sensing using optimized expander graphs, by S. Jafarpour, W. Xu, B. Hassibi, and R. Calderbank. (doi)
  • 03 Dec: Simon Foucart presented Sparse Estimation with Strongly Correlated Variables using Ordered Weighted L1 Regularization, by M. Figueiredo and R. Nowak. (arXiv)