Professional Education


  • Doctor of Philosophy, Stanford University, EE-PHD (2017)
  • Master of Science, Ben Gurion University Of The Negev, Mathematics (2012)
  • Bachelor of Science, Ben Gurion University Of The Negev, Mathematics (2010)
  • Bachelor of Science, Ben Gurion University Of The Negev, Elec. & Comp. Engineering (2010)

Patents


  • Alon Kipnis, Andrea Goldsmith, Yonina C Eldar. "United States Patent 9559714 Analog-to-digital compression", The Board of Trustees of the Leland Stanford Junior University, Jan 31, 2017

Current Research and Scholarly Interests


My recent work explores subtle classification and detection problems involving a vast number of features. My approach, based on multiple hypothesis testing and the Higher Criticism (HC), is particularly appealing when only few features out of possible many are useful while those useful are relatively weak. This approach enables the use of HC and other tests as unsupervised, untrained discriminators: no model is specified for the features hence no learning or tuning is done. This property makes the method incredibly useful in a host of real-world applications. Examples include text classification, detecting mutations in short sequence genomic data, trend prediction in high-dimensional data, social graph analysis, automatic selection of words for topic modeling, and early detection of economic or health crises. As a part of this research, I conducted a massive number of computational experiments to verify the usefulness of the method in these applications.

Another line of my work studies the effect of compression or communication constraints on modern estimation and learning procedures. The disproportional size of datasets compared to computing and communication resources, as well as the wide adoption of cloud computing infrastracture, making such constraints among the most limiting factors in modern data science applications. My work studies fundamental limits of inference and ways to adapt standard methods in statistics and machine learning to the scenario where the data undergoes lossy compression.

My Ph.D. work addressed the effect of data compression on sampling real-world analog data. It provided the first complete characterization of the fundamental trade-off between sampling rate, compression bitrate, and system performance for any digital system processing real-world signal. This tradeoff extends the classical Shannon-Nyquist sampling theorem to the (practical) situation where the samples are quantized or compressed in a lossy manner. In particular, my work shows that, for most signal models, a bitrate constraint imposes a new sampling rate (smaller than Nyquist) above which the signal is optimally represented.

All Publications


  • Two-sample Testing for High-Dimensional Multinomials under Rare/Weak Perturbations Donoho, D. L., Kipnis, A. Department of Statistics. 2020

    Abstract

    Given two samples from possibly different discrete distributions over a common set of size N, consider the problem of testing whether these distributions are identical, vs. the following rare/weak perturbation alternative: the frequencies of N1−β elements are perturbed by r(logN)/2n in the Hellinger distance, where n is the size of each sample. We adapt the Higher Criticism (HC) test to this setting using P-values obtained from N exact binomial tests. We characterize the asymptotic performance of the HC-based test in terms of the sparsity parameter β and the perturbation intensity parameter r. Specifically, we derive a region in the (β,r)-plane where the test asymptotically has maximal power, while having asymptotically no power outside this region. Our analysis distinguishes between the cases of dense (N≫n) and sparse (N≪n) contingency tables. In the dense case, the phase transition curve matches that of an analogous two-sample normal means model.

  • The Compress-and-Estimate Coding Scheme for Gaussian Sources IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS Rini, S., Kipnis, A., Song, R., Goldsmith, A. J. 2019; 18 (9): 4344–56
  • The Distortion-Rate Function of Sampled Wiener Processes IEEE TRANSACTIONS ON INFORMATION THEORY Kipnis, A., Goldsmith, A. J., Eldar, Y. C. 2019; 65 (1): 482–99
  • Gaussian Approximation of Quantization Error for Estimation from Compressed Data Kipnis, A., Reeves, G., IEEE IEEE. 2019: 2029–33
  • Analog-to-Digital Compression A new paradigm for converting signals to bits IEEE SIGNAL PROCESSING MAGAZINE Kipnis, A., Eldar, Y. C., Goldsmith, A. J. 2018; 35 (3): 16–39
  • The Distortion Rate Function of Cyclostationary Gaussian Processes Kipnis, A., Goldsmith, A. J., Eldar, Y. C. IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC. 2018: 3810–24
  • Lossy Compression of Decimated Gaussian Random Walks Murray, G., Kipnis, A., Goldsmith, A. J., IEEE IEEE. 2018
  • Single Letter Formulas for Quantized Compressed Sensing with Gaussian Codebooks Kipnis, A., Reeves, G., Eldar, Y. C., IEEE IEEE. 2018: 71–75
  • Compress-and-Estimate Source Coding for a Vector Gaussian Source Song, R., Rini, S., Kipnis, A., Goldsmith, A. J., IEEE IEEE. 2017: 539–43
  • Coding Theorems for the Compress and Estimate Source Coding Problem Kipnis, A., Rini, S., Goldsmith, A. J., IEEE IEEE. 2017: 2568–72
  • Mean Estimation from Adaptive One-bit Measurements Kipnis, A., Duchi, J. C., IEEE IEEE. 2017: 1000–1007
  • Compressed Sensing under Optimal Quantization Kipnis, A., Reeves, G., Eldar, Y. C., Goldsmith, A. J., IEEE IEEE. 2017: 2148–52
  • Distortion Rate Function of Sub-Nyquist Sampled Gaussian Sources IEEE TRANSACTIONS ON INFORMATION THEORY Kipnis, A., Goldsmith, A. J., Eldar, Y. C., Weissman, T. 2016; 62 (1): 401-429
  • WIENER CHAOS APPROACH TO OPTIMAL PREDICTION NUMERICAL FUNCTIONAL ANALYSIS AND OPTIMIZATION Alpay, D., Kipnis, A. 2015; 36 (10): 1286-1306
  • Distortion Rate Function of Cyclo-Stationary Gaussian Processes IEEE International Symposium on Information Theory (ISIT) Kipnis, A., Goldsmith, A. J. IEEE. 2014: 2834–2838