arXiv subjects

Aleksandr Aravkin

Publications and source records attributed to Aleksandr Aravkin.

43 records · Page 3Linked to original sources

arXiv2017

Distributed Bundle Adjustment

Most methods for Bundle Adjustment (BA) in computer vision are either centralized or operate incrementally. This leads to poor scaling and affects the quality of solution as the number of images grows in large scale structure from motion (SfM). Furthermore, they cannot be used in scenarios where image acquisition and processing must be distributed. We address this problem with a new distributed BA algorithm. Our distributed formulation uses alternating direction method of multipliers (ADMM), and, since each processor sees only a small portion of the data, we show that robust formulations improve performance. We analyze convergence of the proposed algorithm, and illustrate numerical performance, accuracy of the parameter estimates, and scalability of the distributed implementation in the context of synthetic 3D datasets with known camera position and orientation ground truth. The results are comparable to an alternate state-of-the-art centralized bundle adjustment algorithm on synthetic and real 3D reconstruction problems. The runtime of our implementation scales linearly with the number of observed points.

cs.CV↗

arXiv2017

A General Family of Trimmed Estimators for Robust High-dimensional Data Analysis

We consider the problem of robustifying high-dimensional structured estimation. Robust techniques are key in real-world applications which often involve outliers and data corruption. We focus on trimmed versions of structurally regularized M-estimators in the high-dimensional setting, including the popular Least Trimmed Squares estimator, as well as analogous estimators for generalized linear models and graphical models, using possibly non-convex loss functions. We present a general analysis of their statistical convergence rates and consistency, and then take a closer look at the trimmed versions of the Lasso and Graphical Lasso estimators as special cases. On the optimization side, we show how to extend algorithms for M-estimators to fit trimmed variants and provide guarantees on their numerical convergence. The generality and competitive performance of high-dimensional trimmed estimators are illustrated numerically on both simulated and real-world genomics data.

stat.ML↗

arXiv2017

A SMART Stochastic Algorithm for Nonconvex Optimization with Applications to Robust Machine Learning

In this paper, we show how to transform any optimization problem that arises from fitting a machine learning model into one that (1) detects and removes contaminated data from the training set while (2) simultaneously fitting the trimmed model on the uncontaminated data that remains. To solve the resulting nonconvex optimization problem, we introduce a fast stochastic proximal-gradient algorithm that incorporates prior knowledge through nonsmooth regularization. For datasets of size $n$, our approach requires $O(n^{2/3}/\varepsilon)$ gradient evaluations to reach $\varepsilon$-accuracy and, when a certain error bound holds, the complexity improves to $O(κn^{2/3}\log(1/\varepsilon))$. These rates are $n^{1/3}$ times better than those achieved by typical, full gradient methods.

stat.ML↗

arXiv2015

General Optimization Framework for Robust and Regularized 3D Full Waveform Inversion

Scarcity of hydrocarbon resources and high exploration risks motivate the development of high fidelity algorithms and computationally viable approaches to exploratory geophysics. Whereas early approaches considered least-squares minimization, recent developments have emphasized the importance of robust formulations, as well as formulations that allow disciplined encoding of prior information into the inverse problem formulation. The cost of a more flexible optimization framework is a greater computational complexity, as least-squares optimization can be performed using straightforward methods (e.g., steepest descent, Gauss-Newton, L-BFGS), whilst incorporation of robust (non-smooth) penalties requires custom changes that may be difficult to implement in the context of a general seismic inversion workflow. In this study, we propose a generic, flexible optimization framework capable of incorporating a broad range of noise models, forward models, regularizers, and reparametrization transforms. This framework covers seamlessly robust noise models (such as Huber and Student's $t$), as well as sparse regularizers, projected constraints, and Total Variation regularization. The proposed framework is also expandable --- we explain the adjustments that are required for any new formulation to be included. Lastly, we conclude with few numerical examples demonstrating the versatility of the formulation.

math.OC↗

arXiv2014

A variational approach to stable principal component pursuit

We introduce a new convex formulation for stable principal component pursuit (SPCP) to decompose noisy signals into low-rank and sparse representations. For numerical solutions of our SPCP formulation, we first develop a convex variational framework and then accelerate it with quasi-Newton methods. We show, via synthetic and real data experiments, that our approach offers advantages over the classical SPCP formulations in scalability and practical parameter selection.

math.OC↗

arXiv2013

Iterative Log Thresholding

Sparse reconstruction approaches using the re-weighted l1-penalty have been shown, both empirically and theoretically, to provide a significant improvement in recovering sparse signals in comparison to the l1-relaxation. However, numerical optimization of such penalties involves solving problems with l1-norms in the objective many times. Using the direct link of reweighted l1-penalties to the concave log-regularizer for sparsity, we derive a simple prox-like algorithm for the log-regularized formulation. The proximal splitting step of the algorithm has a closed form solution, and we call the algorithm 'log-thresholding' in analogy to soft thresholding for the l1-penalty. We establish convergence results, and demonstrate that log-thresholding provides more accurate sparse reconstructions compared to both soft and hard thresholding. Furthermore, the approach can be directly extended to optimization over matrices with penalty for rank (i.e. the nuclear norm penalty and its re-weigthed version), where we suggest a singular-value log-thresholding approach.

stat.ML↗

arXiv2012

Robust inversion via semistochastic dimensionality reduction

We consider a class of inverse problems where it is possible to aggregate the results of multiple experiments. This class includes problems where the forward model is the solution operator to linear ODEs or PDEs. The tremendous size of such problems motivates dimensionality reduction techniques based on randomly mixing experiments. These techniques break down, however, when robust data-fitting formulations are used, which are essential in cases of missing data, unusually large errors, and systematic features in the data unexplained by the forward model. We survey robust methods within a statistical framework, and propose a semistochastic optimization approach that allows dimensionality reduction. The efficacy of the methods are demonstrated for a large-scale seismic inverse problem using the robust Student's t-distribution, where a useful synthetic velocity model is recovered in the extreme scenario of 60% data missing at random. The semistochastic approach achieves this recovery using 20% of the effort required by a direct robust approach.

cs.CE↗