Ilija Bogunovic
École Polytechnique Fédérale de Lausanne
Network
Latest external collaboration on country level. Dive into details by clicking on the dots.
Publication
Featured researches published by Ilija Bogunovic.
IEEE Journal of Selected Topics in Signal Processing | 2016
Luca Baldassarre; Yen-Huan Li; Jonathan Scarlett; Baran Gözcü; Ilija Bogunovic; Volkan Cevher
The problem of recovering a structured signal x ∈ C<sup>p</sup> from a set of dimensionality-reduced linear measurements b = Ax arises in a variety of applications, such as medical imaging, spectroscopy, Fourier optics, and computerized tomography. Due to computational and storage complexity or physical constraints imposed by the problem, the measurement matrix A ∈ C<sup>n×p</sup> is often of the form A = P<sub>Ω</sub>Ψ for some orthonormal basis matrix Ψ ∈ C<sup>P×P</sup> and subsampling operator P<sub>Ω</sub> : C<sup>p</sup> → C<sup>n</sup> that selects the rows indexed by Ω. This raises the fundamental question of how best to choose the index set Ω in order to optimize the recovery performance. Previous approaches to addressing this question rely on nonuniform random subsampling using application-specific knowledge of the structure of x. In this paper, we instead take a principled learning-based approach in which affixed index set is chosen based on a set of training signals x<sub>1</sub>, . . . , x<sub>m</sub>. We formulate combinatorial optimization problems seeking to maximize the energy captured in these signals in an average-case or worst-case sense, and we show that these can be efficiently solved either exactly or approximately via the identification of modularity and submodularity structures. We provide both deterministic and statistical theoretical guarantees showing how the resulting measurement matrices perform on signals differing from the training signals, and we provide numerical examples showing our approach to be effective on a variety of data sets.
international conference on acoustics, speech, and signal processing | 2015
Ilija Bogunovic; Volkan Cevher; Jarvis D. Haupt; Jonathan Scarlett
We study the problem of actively learning a multi-index function of the form f(x) = g<sub>0</sub>(A<sub>0</sub>x) from its point evaluations, where A<sub>0</sub> ∈ ℝ<sup>k×d</sup> with k ≫ d. We build on the assumptions and techniques of an existing approach based on low-rank matrix recovery (Tyagi and Cevher, 2012). Specifically, by introducing an additional self- concordant like assumption on g0 and adapting the sampling scheme and its analysis accordingly, we provide a bound on the sampling complexity with a weaker dependence on d in the presence of additive Gaussian sampling noise. For example, under natural assumptions on certain other parameters, the dependence decreases from O(d<sup>3/2</sup>) to O(d<sup>¾</sup>).
international conference on machine learning | 2014
Adish Singla; Ilija Bogunovic; Gábor Bartók; Amin Karbasi; Andreas Krause
neural information processing systems | 2013
Adish Singla; Ilija Bogunovic; Gábor Bartók; Amin Karbasi; Andreas Krause
international conference on artificial intelligence and statistics | 2016
Ilija Bogunovic; Jonathan Scarlett; Volkan Cevher
international conference on machine learning | 2017
Ilija Bogunovic; Slobodan Mitrovic; Jonathan Scarlett; Volkan Cevher
neural information processing systems | 2018
Ilija Bogunovic; Jonathan Scarlett; Stefanie Jegelka; Volkan Cevher
international conference on artificial intelligence and statistics | 2018
Paul Thierry Yves Rolland; Jonathan Scarlett; Ilija Bogunovic; Volkan Cevher
international conference on artificial intelligence and statistics | 2018
Ilija Bogunovic; Junyao Zhao; Volkan Cevher
neural information processing systems | 2017
Slobodan Mitrovic; Ilija Bogunovic; Ashkan Norouzi-Fard; Jakub Tarnawski; Volkan Cevher