Publications by Anastasios Kyrillidis

A single-phase, proximal path-following framework

Volkan Cevher, Anastasios Kyrillidis, Quoc Tran Dinh

We propose a new proximal, path-following framework for a class of---possibly non-smooth---constrained convex problems. We consider settings where the non-smooth part is endowed with a proximity operator, and the constraint set is equipped with a self-conc ...

INFORMS2018

A Non-Euclidean Gradient Descent Framework for Non-Convex Matrix Factorization

Volkan Cevher, Anastasios Kyrillidis, Rabeeh Karimi Mahabadi, Ya-Ping Hsieh, Yu-Chun Kao

We study convex optimization problems that feature low-rank matrix solutions. In such scenarios, non-convex methods offer significant advantages over convex methods due to their lower space complexity as well as faster convergence speed. Moreover, many of ...

2018

Group-Sparse Model Selection: Hardness and Relaxations

Volkan Cevher, Anastasios Kyrillidis, Luca Baldassarre, Nirav Bhan

Group-based sparsity models are proven instrumental in linear regression problems for recovering signals from much fewer measurements than standard compressive sensing. The main promise of these models is the recovery of ``interpretable" signals along with ...

Institute of Electrical and Electronics Engineers2016

Convex block-sparse linear regression with expanders - provably

Volkan Cevher, Anastasios Kyrillidis, Luca Baldassarre, Bubacarr Bah, Quoc Tran Dinh

Sparse matrices are favorable objects in machine learning and optimization. When such matrices are used, in place of dense ones, the overall complexity requirements in optimization can be significantly reduced in practice, both in terms of space and run-ti ...

2016

Composite Self-Concordant Minimization

Volkan Cevher, Anastasios Kyrillidis, Quoc Tran Dinh

We propose a variable metric framework for minimizing the sum of a self-concordant function and a possibly non-smooth convex function endowed with a computable proximal operator. We theoretically establish the convergence of our framework without relying o ...

Massachusetts Institute of Technology Press2015

Rigorous optimization recipes for sparse and low rank inverse problems with applications in data sciences

Anastasios Kyrillidis

Many natural and man-made signals can be described as having a few degrees of freedom relative to their size due to natural parameterizations or constraints; examples include bandlimited signals, collections of signals observed from multiple viewpoints in ...

EPFL2014

Matrix Recipes for Hard Thresholding Methods

Volkan Cevher, Anastasios Kyrillidis

In this paper, we present and analyze a new set of low-rank recovery algorithms for linear inverse problems within the class of hard thresholding methods. We provide strategies on how to set up these algorithms via basic ingredients for different configura ...

Springer Verlag2014

Scalable sparse covariance estimation via self-concordance

Volkan Cevher, Anastasios Kyrillidis, Rabeeh Karimi Mahabadi, Quoc Tran Dinh

We consider the class of convex minimization problems, composed of a self-concordant function, such as the logdet metric, a convex data fidelity term h and, a regularizing -- possibly non-smooth -- function g. This type of problems have recently attracted ...

2014

An inexact proximal path-following algorithm for constrained convex minimization

Volkan Cevher, Anastasios Kyrillidis, Quoc Tran Dinh

Many scientific and engineering applications feature large-scale non-smooth convex minimization problems over convex sets. In this paper, we address an important instance of this broad class where we assume that the non-smooth objective is equipped with a ...

Siam Publications2014

Approximate Matrix Multiplication with Application to Linear Embeddings

Anastasios Kyrillidis, Michail Vlachos

In this paper, we study the problem of approximately computing the product of two real matrices. In particular, we analyze a dimensionality-reduction-based approximation algorithm due to Sarlos [1], introducing the notion of nuclear rank as the ratio of th ...

Ieee2014

Fixed-Rank Rayleigh Quotient Maximization by an MPSK Sequence

Anastasios Kyrillidis

Certain optimization problems in communication systems, such as limited-feedback constant-envelope beamforming or noncoherent M-ary phase-shift keying (MPSK) sequence detection, result in the maximization of a fixed-rank positive semidefinite quadratic for ...

Institute of Electrical and Electronics Engineers2014

A proximal Newton framework for composite minimization: Graph learning without Cholesky decompositions and matrix inversions

Volkan Cevher, Anastasios Kyrillidis, Quoc Tran Dinh

We propose an algorithmic framework for convex minimization problems of a composite function with two terms: a self-concordant function and a possibly nonsmooth regularization term. Our method is a new proximal Newton algorithm that features a local quadra ...

2013

Sparse projections onto the simplex

Volkan Cevher, Anastasios Kyrillidis, Christoph Koch

Most learning methods with rank or sparsity constraints use convex relaxations, which lead to optimization with the nuclear norm or the`1-norm. However, several important learning applications cannot benet from this approach as they feature these convex no ...

JMLR W&CP2013

Fast Proximal algorithms for Self-concordant function minimization with application to sparse graph selection

Volkan Cevher, Anastasios Kyrillidis

The convex

\ell_1

-regularized

\log\det

divergence criterion has been shown to produce theoretically consistent graph learning. However, this objective function is challenging since the

\ell_1

-regularization is nonsmooth, the

\log\det

objective is n ...

2013

Fast Proximal Algorithms For Self-Concordant Function Minimization With Application To Sparse Graph Selection

Volkan Cevher, Anastasios Kyrillidis

The convex l(1)-regularized log det divergence criterion has been shown to produce theoretically consistent graph learning. However, this objective function is challenging since the l(1)-regularization is nonsmooth, the log det objective is not globally Li ...

IEEE2013

Multi-Way Compressed Sensing for Sparse Low-Rank Tensors

Anastasios Kyrillidis

For linear models, compressed sensing theory and methods enable recovery of sparse signals of interest from few measurements-in the order of the number of nonzero entries as opposed to the length of the signal of interest. Results of similar flavor have mo ...

IEEE Institute of Electrical and Electronics Engineers2012

Hard Thresholding with Norm Constraints

Gilles Puy, Volkan Cevher, Anastasios Kyrillidis

We introduce a new sparse recovery paradigm, called Normed Pursuits, where efficient algorithms from combinatorial and convex optimization interface for interpretable and model-based solutions. Synthetic and real data experiments illustrate that Normed Pur ...

Ieee2012

Combinatorial Selection and Least Absolute Shrinkage via the CLASH Algorithm

Volkan Cevher, Anastasios Kyrillidis

The least absolute shrinkage and selection operator (LASSO) for linear regression exploits the geometric interplay of the

\ell_2

-data error objective and the

\ell_1

-norm constraint to arbitrarily select sparse models. Guiding this uninformed selection ...

IEEE2012

Matrix ALPS: Accelerated Low Rank and Sparse Matrix Reconstruction

Volkan Cevher, Anastasios Kyrillidis

We propose Matrix ALPS for recovering a sparse plus low-rank decomposition of a matrix given its corrupted and incomplete linear measurements. Our approach is a first-order projected gradient method over non-convex sets, and it exploits a well-known memory ...

IEEE2012

Rank-Deficient Quadratic-Form Maximization Over M-Phase Alphabet: Polynomial-Complexity Solvability And Algorithmic Developments

Anastasios Kyrillidis

The maximization of a positive (semi) definite complex quadratic form over a finite alphabet is NP-hard and achieved through exhaustive search when the form has full rank. However, if the form is rank-deficient, the optimal solution can be computed with on ...

IEEE2011