Aleksandr Aravkin
Curriculum Vitae
Contact
Information
IBM T. J. Watson Research Center,
Voice: (206) 459-4734
Yorktown Heights, NY 10598
E-mail: saravkin@us.ibm.com
Research Focus
I work in large scale optimization, machine learning and data science. My interests in-
clude sparse/stochastic/PDE constrained optimization, convex/variational analysis, robust
statistics, deep learning, dynamic systems, signal processing, and 3D image reconstruction.
Education
Ph.D., Mathematics, 2010, University of Washington, Seattle, WA.
- Thesis: Robust methods for Kalman filtering/smoothing and Bundle Adjustment.
University of Washington, 2010, advised by J.V. Burke.
M.S., Statistics, 2010, University of Washington, Seattle, WA.
B.S., Mathematics and Computer Science, 2004, University of Washington, Seattle, WA.
Appointments
Columbia University, Computer Science & IEOR
01/2014-Now
:
Adjunct Professor.
Teach graduate courses in machine learning and high dimensional data analysis,
with focus on optimization, scientific computing, and statistical modeling.
IBM Thomas J. Watson Research Center
01/2013-Now
:
Research Staff Member.
Engage in pure and applied research in industry, building analytics for
renewable energy utilities, seismic imaging and inversion, and a range
of applications that depend on time series data.
University of British Columbia, Computer Science and Earth/Ocean Sciences.
2010-2012
:
Postdoctoral Fellow.
Duties included research, supervising students, and active engagement
with researchers and industrial sponsors in geophysics.
University of Washington
2004-2010
:
Instructor, Research Assistant, Teaching Assistant, Dept. of Mathematics.
NASA Ames Research Center
Summer 2007-2009
:
Educational Associate, Space Grant Intern, Intelligent Robotics Group.
Honors and
Awards
IBM Nominee for Singapore Global Young Scientist Summit, 2014.
12 IBM Invention Achievement Awards, 2013-2014.
NSF Vertical Integration Grant for Research and Education (VIGRE) Fellow, 2008.
NASA Space Grant Recipient, 2007 and 2008.
Outstanding Winner of Mathematical Contest in Modeling (MCM), 2004.
University of Washington Mathematics Department Academic Excellence Award, 2004.
1
Teaching
Experience
Columbia University, Computer Science, 2014:
• Teaching high dimensional data analysis (COMSE 6998), a graduate course in statis-
tical analysis and algorithms for sparse regression, low rank recovery, and extensions.
• Taught Advanced Machine Learning (COMSE 4772), a graduate course for the focus-
ing on optimization in machine learning, Spring 2013.
University of British Columbia, 2010-2012: Co-supervised masters and PhD students in
Earth and Ocean Sciences and Computer Science, engaging them on topics in mathematical
modeling, statistics, optimization theory and algorithm design.
University of Washington, 2004-2010:
• Taught discrete math modeling (UW Math 381), covering linear programming, net-
work optimization, statistics, stochastic processes, Monte Carlo methods, and Markov
chains. Supervised teams in modeling and solving real-world problems, 2010.
• Taught differential equations (UW Math 307), 2009.
• Teaching mentor for fellow graduate students. Activity included observing sections,
giving feedback on teaching, and helping resolve issues for new instructors, 2009.
• Teaching assistant: taught quiz sections, graded, and held office hours for Algebra and
Calculus for Biology/Business/Economics, math modeling, linear programming, and
combinatorics, 2004-2009.
In Review
A.Y. Aravkin, J.V. Burke, and G. Pillonetto. Kalman smoothing and block tridiagonal
systems: new connections and stability results. (15 pages)
A.Y. Aravkin, J.V. Burke, G. Pillonetto. New stable spline estimators for robust, sparse
and inequality constrained linear system identification. (17 pages)
K.N. Ramamurthy, A.Y. Aravkin, J. J. Thiagarajan.
Beyond L2-Loss Functions for
Learning Sparse Models. (10 pages)
C-C Lin, S. U. Pankanti, K. N. Ramamurthy, and A.Y. Aravkin. Content and Perspective
Preserving Stitching. (9 pages)
G. Bottegal, A.Y. Aravkin, H. Hjalmarsson and G. Pillonetto. Robust EM kernel-based
methods for linear system identification. (15 pages)
To appear
A.Y. Aravkin, A. Kambadur, A.C. Lozano, and R. Luss. Orthogonal Matching Pursuit for
Sparse Quantile Regression. arXiv:1402.4624, to appear in IEEE International Conference
on Data Mining (ICDM) 2014. (10 pages)
A.Y. Aravkin, S. Becker, V. Cevher, and P. Olsen. A variational approach to stable princi-
pal component pursuit.
arXiv:1406.1089, to appear in Uncertainty in Artificial Intelligence
(UAI) Proceedings, 2014 (10 pages).
A.Y. Aravkin, B.M. Bell, J.V. Burke, and G. Pillonetto. The connection between Bayesian
estimation of a Gaussian random field and RKHS. arXiv:1301.5288, to appear in IEEE
Transactions on Neural Networks and Learning Systems (TNNLS), 2014 (8 pages).
A.Y. Aravkin, K.N. Ramamurthy, and G. Pillonetto. Kalman Smoothing With Persistent
Nuisance Parameters. To appear in proceedings for Machine Learning for Signal Processing
(MLSP), 2014 (6 pages).
G. Pillonetto and A.Y. Aravkin. A New Kernel-Based Approach For Identification Of
Time-Varying Linear Systems. To appear in proceedings for Machine Learning for Signal
Processing (MLSP), 2014 (6 pages).
2
O. Tripp, S. Guarnieri, M. Pistoia and A. Aravkin.
Aletheia: Improving the Usability of
Static Security Analysis.
To appear in proceedings of ACM CSS 2024 (10 pages).
A.Y. Aravkin and J.V. Burke. Smoothing dynamic systems with state-dependent covari-
ance matrices.
arXiv:1211.4601, to appear in the Conference on Decision and Control
(CDC), 2014 (6 pages).
Journal
Publications
A.Y. Aravkin, R. Kumar, H. Mansour, B. Recht, and F.J. Herrmann. Fast methods for
denoising matrix completion formulations, with applications to robust seismic data interpo-
lation. SIAM J. Sci. Comput., 36(5):S237-S266, 2014.
A.Y. Aravkin, J.V. Burke, G. Pillonetto. Robust and trend following Student’s t Kalman
smoothers. SIAM J. Control Optim., 52(5):2891-2916, 2014.
A.Y. Aravkin, J.V. Burke, and G. Pillonetto. Optimization viewpoint on Kalman smooth-
ing, with applications to robust and sparse estimation. In Compressed Sensing and Sparse
Filtering, pages 237-280. Springer, 2014.
A.Y. Aravkin, J.V. Burke, A. Chiuso, and G. Pillonetto. Convex vs. nonconvex approaches
for sparse estimation: the mean squared properties of ARD and GLasso. Journal of Machine
Learning Research (JMLR), Volume 15, pages 217-252, 2014.
I. Iskander, R. Gamaleldin, S. El Houchi, A. El Shenawy, I. Seoud, N. El Gharbawi, H.
Abou-Youssef, A.Y. Aravkin, and R. Wennberg, Serum Bilirubin and Bilirubin/Albumin
Ratio as Predictors of Bilirubin Encephalopathy. Pediatrics, DOI: 10.1542/peds.2013-1764,
2014.
E. Khan, A.Y. Aravkin, M.P. Friedlander, and M. Seeger. Fast dual variational inference
for non-conjugate latent Gaussian models.
JMLR W&CP 28(3):951-959, 2013.
A.Y. Aravkin, J.V. Burke, and G. Pillonetto. Sparse/Robust Estimation and Kalman
Smoothing with Nonsmooth Log-Concave Densities: Modeling, Computation, and Theory.
Journal of Machine Learning Research (JMLR), Volume 3, pages 2689-2728, September
2013.
A.Y. Aravkin, J.V. Burke, M.P. Friedlander. Variational Properties of Value Functions.
SIAM Journal of Optimization (SIOPT), Vol. 23, No. 3, pp. 1689-1717, August 2013.
C.L. Wang, M.J. Eissa, J.V. Rogers, A.Y. Aravkin, B.A. Porter, J.D. Beatty. (18)F-FDG
PET/CT-Positive Internal Mammary Lymph Nodes: Pathologic Correlation by Ultrasound-
Guided Fine-Needle Aspiration and Assessment of Associated Risk Factors. AJR Am J
Roentgenol. 2013 May; 200(5): 1138-1144.
A.Y. Aravkin, T. van Leeuwen. Estimating Nuisance Parameters in Inverse Problems,
Inverse Problems, 28(11):115016, October 2012, 17 pages.
A.Y. Aravkin, M.P. Friedlander, F.J. Herrmann, and T. van Leeuwen. Robust Inversion,
Dimensionality Reduction, and Randomized Sampling, Mathematical Programming, 2012,
Volume 134, Number 1, Pages 101-125.
X. Li, A.Y. Aravkin, T. van Leeuwen, and F.J. Herrmann, Fast randomized full-waveform
inversion with compressive sensing, Geophysics 77(3), A13-A17, 2012.
T. van Leeuwen, A.Y. Aravkin, and F.J. Herrmann, Seismic waveform inversion by
stochastic optimization, International Journal of Geophysics, vol. 2011, 18 pages.
Aravkin, A.Y., Bell, B.M., Burke, J.V., Pillonetto, G., (2009) An
1
-Laplace Robust
3
Kalman Smoother, IEEE Transactions of Automatic Control (TAC), 56(12):2898-2911, 2011.
R. Gammeldin, I. Iskander, I. Seoud, H. Aboraya, A.Y. Aravkin, P. Sampson, R. Wennberg,
Risk Factors of Neurotoxicity in Newborns with Severe Neonatal Hyperbilirubinemia. Pedi-
atrics 128(4), 2011, e925-e931.
C. Wang, L. MacDonald, J. Rogers, A.Y. Aravkin, D. Haseley, J. Beatty, Positron emis-
sion mammography: correlation of estrogen receptor, progresterone receptor, and human
epidermal growth factor receptor 2 status and 18F-FDG.
AJR Am J Roentgenol. 2011
Aug; 197(2): W247-55.
R. Wennberg, C. Ahlfors and A.Y. Aravkin, Intervention guidelines for neonatal hyper-
bilirubinemia: an evidence based quagmire. Curr. Pharm. Design 2009; 15(25):2939-45.
Conference
Proceedings
A.Y. Aravkin, A. Choromanska, T. Jebara and D. Kanevsky. Semistochastic Quadratic
Bound Methods.
arXiv:1309.1369, presented at the International Conference on Learning
Representations (ICLR) Workshop, 2014 (11 pages).
R. Kumar, A.Y. Aravkin, E. Esser, H. Mansour and F.J. Herrmann.
SVD-free low-
rank matrix factorization : wavefield reconstruction via jittered subsampling and reciprocity.
Proceedings of European Association of Geoscientists and Engineers (EAGE), January 2014
(5 pages).
G. Bottegal, A.Y. Aravkin, H. Hjalmarsson and G. Pillonetto. Outlier robust system
identification: a Bayesian kernel-based approach.
Proceedings of International Federation
of Automatic Control (IFAC), Volume 19, Part 1, 2014 (6 pages).
D. Malioutov, A.Y. Aravkin. Iterative log thresholding. In Acoustics, Speech and Signal
Processing (ICASSP), 2014 IEEE International Conference on, p. 7198-7202.
T.N. Sainath, L. Horesh, B. Kingsbury, A.Y. Aravkin, B. Ramabhadran. Accelerating
Hessian-free optimization for deep neural networks by implicit preconditioning and sampling.
arXiv:1309.1508, in proceedings of IEEE Automatic Speech Recognition and Understanding
(ASRU) 2013 (6 pages).
T.N. Sainath, B. Kingsbury, A. Mohamed, G.E. Dahl, G. Saon, H. Soltau, T. Beran,
A.Y. Aravkin, B. Ramabhadran. Improvements for deep convolutional neural networks
for LVCSR. arXiv:1309.1501, in proceedings of IEEE Automatic Speech Recognition and
Understanding (ASRU) 2013 (6 pages).
A.Y. Aravkin, J.V. Burke and G. Pillonetto. Linear system identification using stable
spline kernels and PLQ penalties. Decision and Control (CDC), 2013 IEEE 52nd Annual
Conference on, p. 5168-5173.
R. Kumar, H. Mansour, A.Y. Aravkin, and F.J. Herrmann. Reconstruction of seismic
wavefields via low-rank matrix factorization in the HSS matrix representation. Proceedings
of the Society of Exploration Geophysicists (SEG), April 2013 (5 pages).
A.Y. Aravkin, T. van Leeuwen and N. Tu. Sparse seismic imaging using variable pro-
jection. In Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International
Conference on, p. 2065-2069.
T. van Leeuwen, A.Y. Aravkin, H. Calandra, and F.J. Herrmann. Which domain for
robust full waveform inversion? In Proceedings of the EAGE, January 2013 (5 pages).
A.Y. Aravkin, J.V. Burke and G. Pillonetto. Nonsmooth regression and state estimation
using piecewise quadratic log-concave densities. In Proceedings of IEEE Conf. Decision and
4
Control (CDC) 2012, p. 4101-4106.
F.J. Herrmann, X. Li, A.Y. Aravkin, and T. van Leeuwen. A modified, sparsity promoting,
Gauss-Newton algorithm for seismic waveform inversion. In Proc. International Society for
Optics and Photonics (SPIE) 8138, 81380V (2011) (14 pages).
A.Y. Aravkin, J.V. Burke, A. Chiuso, G. Pillonetto, Convex vs. nonconvex approaches
for sparse estimation: Lasso, Multiple Kernel Learning and Hyperparameter Lasso.
In
Proceedings of IEEE Conference on Decision and Conrol (CDC) Proceedings, 2011, 8 pages.
A.Y. Aravkin, T. van Leeuwen, K. Bube and F.J. Herrmann.
On non-uniqueness of
the Student’s t formulation for linear inverse problems.
In Proceedings of the Society of
Exploration Geophysicists (SEG), 2012, 4 pages.
A.Y. Aravkin, M. Styer, Z. Moratto, A. Nefian, and M. Broxton.
Student’s t robust
bundle adjustment algorithm.
Image Processing (ICIP), 2012 19th IEEE International
Conference on, p. 1757-1760.
A.Y. Aravkin, J.V. Burke, A. Chiuso and G. Pillonetto.
On the MSE Properties of
Empirical Bayes Methods for Sparse Estimation. in System Identification, Volume 16 (1),
p. 965-970, 2012.
A.Y. Aravkin, J.V. Burke, and G. Pillonetto. Robust and Trend-following Kalman smoothers
using Student’s t. in System Identification, Volume 16 (1) p. 1215–1220, 2012.
A.Y. Aravkin, J.V. Burke, and G. Pillonetto. A statistical and computational theory for
robust and sparse Kalman smoothing. in System Identification, Volume 16 (1), 894–899,
2012.
A.Y. Aravkin, J.V. Burke, A. Chiuso and G. Pillonetto. On the estimation of hyperparam-
eters for Empirical Bayes estimators: Maximum Marginal Likelihood vs Minimum MSE. in
System Identification, Volume 16 (1), p. 125-130, 2012.
A.Y. Aravkin, X. Li, and F.J. Herrmann. Fast seismic imaging for marine data. Acous-
tics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on, p.
2517-2520.
A.Y. Aravkin, M.P. Friedlander, and T. van Leeuwen. Robust inversion via semistochastic
dimensionality reduction. Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE
International Conference on, p. 5245-5248.
A.Y. Aravkin, T. van Leeuwen, H. Calandra, and F.J. Herrmann. Source estimation for
frequency-domain FWI with robust penalties. In Proceedings of EAGE 2012 (5 pages).
X. Li, A.Y. Aravkin, T. van Leeuwen, F.J. Herrmann, Modified Gauss-Newton with Sparse
Updates. Proc. Brazilian Geophysical Society (SBGF) 2011, (5 pages).
A.Y. Aravkin, T. van Leeuwen, F.J. Herrmann, Robust full-waveform inversion using the
Student’s t-distribution, SEG Expanded Abstracts 30, 2669 (2011) (5 pages).
A.Y. Aravkin, T. van Leeuwen, J.V. Burke, F.J. Herrmann, A nonlinear sparsity pro-
moting formulation and algorithm for full waveform inversion, EAGE Expanded Abstracts
2011 (5 pages).
X. Li, A.Y. Aravkin, T. van Leeuwen, F.J. Herrmann, Full-waveform inversion with ran-
domized L1 recovery for the model updates. EAGE Expanded Abstracts 2011 (5 pages).
A.Y. Aravkin, B. Bell, J.V. Burke, G. Pillonetto, Learning Using State Space Kernel
5
Machines, IFAC World Congress, Volume 18, Part 1 2011 (7 pages).
G.Pillonetto, A.Y. Aravkin, S. Carpin, The unconstrained and inequality constrained mov-
ing horizon approach to robot localization. Proceedings of the 2010 IEEE/RSJ International
Conference on Intelligent Robots and Systems, 3830-3835.
Contributed
Software
Software package for robust and constrained Kalman smoothing.
See COIN-OR, https://projects.coin-or.org/CoinBazaar/wiki/Projects/ckbs.
General interior point solver for a broad range of signal processing and machine learning
formulations.
See https://github.com/saravkin/ipsolve.
Extended sparse optimization package SPG
1
to handle more general models and applica-
tions, including robust methods and fitting nuisance parameters.
See https://github.com/saravkin/spgl1.
Developed and implemented robust Bundle Adjustment methods for NASA’s Vision Work-
bench package. See https://github.com/visionworkbench.
Personal
Citizenship: USA.
Languages: English, Russian, French.
6
Dostları ilə paylaş: |