Ivan Dokmanić

Ivan Dokmanić

Auto-generated profile, based on publicly available information

Ivan Dokmanić

Department of Mathematics and Computer Science, University of Basel, Basel, Switzerland

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 May 2022 in IEEE Transactions on Computational Imaging
Verified
Journal Article
Open access
TLTianlin Liu
Tianlin Liu
ACAnadi Chaman
Anadi Chaman
DBDavid Belius
David Belius
IDIvan Dokmanic
Ivan Dokmanic
Convolutional neural networks (CNNs) have been tremendously successful in solving imaging inverse problems. To understand their success, an effective strategy is to construct simpler and mathematically more tractable convolutional sparse coding (CSC) models that share essential ingredients with CNNs. Existing CSC methods, however, underperform leading CNNs in challenging inverse problems. We hypothesize that the performance gap may be attributed in part to how they process images at different spatial scales: While many CNNs use multiscale feature representations, existing CSC models mostly rely on single-scale dictionaries. To close the performance gap, we thus propose a multiscale convolutional dictionary structure. The proposed dictionary structure is derived from the U-Net, arguably the most versatile and widely used CNN for image-to-image learning problems. We show that incorporating the proposed multiscale dictionary in an otherwise standard CSC framework yields performance competitive with state-of-the-art CNNs across a range of challenging inverse problems including CT and MRI reconstruction. Our work thus demonstrates the effectiveness and scalability of the multiscale CSC approach in solving challenging inverse problems.

1 April 2019 in IEEE Signal Processing Letters
Verified
Journal Article
Open access
IDIvan Dokmanic
Ivan Dokmanic
A recent unlabeled sampling result by Unnikrishnan, Haghighatshoar and Vetterli states that with probability one over iid Gaussian matrices $A$ , any $x$ can be uniquely recovered from an unknown permutation of $y = A x$ as soon as $A$ has at least twice as many rows as columns. We show that this condition on $A$ implies something much stronger: that an unknown vector $x$ can be recovered from measurements $y = T A x$ , when the unknown $T$ belongs to an arbitrary set of invertible, diagonalizable linear transformations $\mathcal{T}$ . The set $\mathcal{T}$ can be finite or countably infinite. When it is the set of $m \times m$ permutation matrices, we have the classical unlabeled sampling problem. We show that for almost all $A$ with at least twice as many rows as columns, all $x$ can be recovered either uniquely, or up to a scale depending on $\mathcal{T}$ , and that the condition on the size of $A$ is necessary. Our proof is based on vector space geometry. Specializing to permutations we obtain a simplified proof of the uniqueness result of Unnikrishnan, Haghighatshoar and Vetterli. In this letter we are only concerned with uniqueness; stability and algorithms are left for future work.

30 October 2019 in Seismological Research Letters
Verified
Journal Article
Closed access
PDF file
HFHongjian Fang
Hongjian Fang
RHRobert D. van der Hilst
Robert D. van der Hilst
MHMaarten V. de Hoop
Maarten V. de Hoop
KKKonik Kothari
Konik Kothari
SGSidharth Gupta
Sidharth Gupta
IDIvan Dokmanić
Ivan Dokmanić
Ill‐posed seismic inverse problems are often solved using Tikhonov‐type regularization, that is, incorporation of damping and smoothing to obtain stable results. This typically results in overly smooth models, poor amplitude resolution, and a difficult choice between plausible models. Recognizing that the average of parameters can be better constrained than individual parameters, we propose a seismic tomography method that stabilizes the inverse problem by projecting the original high‐dimension model space onto random low‐dimension subspaces and then infers the high‐dimensional solution from combinations of such subspaces. The subspaces are formed by functions constant in Poisson Voronoi cells, which can be viewed as the mean of parameters near a certain location. The low‐dimensional problems are better constrained, and image reconstruction of the subspaces does not require explicit regularization. Moreover, the low‐dimension subspaces can be recovered by subsets of the whole dataset, which increases efficiency and offers opportunities to mitigate uneven sampling of the model space. The final (high‐dimension) model is then obtained from the low‐dimension images in different subspaces either by solving another normal equation or simply by averaging the low‐dimension images. Importantly, model uncertainty can be obtained directly from images in different subspaces. Synthetic tests show that our method outperforms conventional methods both in terms of geometry and amplitude recovery. The application to southern California plate boundary region also validates the robustness of our method by imaging geologically consistent features as well as strong along‐strike variations of San Jacinto fault that are not clearly seen using conventional methods.

11 December 2019 in IEEE Transactions on Signal Processing
Verified
Journal Article
Open access
PTPuoya Tabaghi
Puoya Tabaghi
IDIvan Dokmanic
Ivan Dokmanic
MVMartin Vetterli
Martin Vetterli
Euclidean distance matrices (EDMs) are a major tool for localization from distances, with applications ranging from protein structure determination to global positioning and manifold learning. They are, however, static objects which serve to localize points from a snapshot of distances. If the objects move, one expects to do better by modeling the motion. In this paper, we introduce Kinetic Euclidean Distance Matrices (KEDMs)---a new kind of time-dependent distance matrices that incorporate motion. The entries of KEDMs become functions of time, the squared time-varying distances. We study two smooth trajectory models---polynomial and bandlimited trajectories---and show that these trajectories can be reconstructed from incomplete, noisy distance observations, scattered over multiple time instants. Our main contribution is a semidefinite relaxation (SDR), inspired by SDRs for static EDMs. Similarly to the static case, the SDR is followed by a spectral factorization step; however, because spectral factorization of polynomial matrices is more challenging than for constant matrices, we propose a new factorization method that uses anchor measurements. Extensive numerical experiments show that KEDMs and the new semidefinite relaxation accurately reconstruct trajectories from noisy, incomplete distance data and that, in fact, motion improves rather than degrades localization if properly modeled. This makes KEDMs a promising tool for problems in geometry of dynamic points sets.

5 March 2021 in IEEE Transactions on Signal Processing
Verified
Journal Article
Open access
SHShuai Huang
Shuai Huang
IDIvan Dokmanic
Ivan Dokmanic
We address the problem of reconstructing a set of points on a line or a loop from their unassigned noisy pairwise distances. When the points lie on a line, the problem is known as the turnpike; when they are on a loop, it is known as the beltway. We approximate the problem by discretizing the domain and representing the $N$ points via an $N$ -hot encoding, which is a density supported on the discretized domain. We show how the distance distribution is then simply a collection of quadratic functionals of this density and propose to recover the point locations so that the estimated distance distribution matches the measured distance distribution. This can be cast as a constrained nonconvex optimization problem which we solve using projected gradient descent with a suitable spectral initializer. We derive conditions under which the proposed distance distribution matching approach locally converges to a global optimizer at a linear rate. Compared to the conventional backtracking approach, our method jointly reconstructs all the point locations and is robust to noise in the measurements. We substantiate these claims with state-of-the-art performance across a number of numerical experiments. Our method is the first practical approach to solve the large-scale noisy beltway problem where the points lie on a loop.

22 February 2023 in IEEE Transactions on Signal Processing
Verified
Journal Article
Open access
DBDalia El Badawy
Dalia El Badawy
VLViktor Larsson
Viktor Larsson
MPMarc Pollefeys
Marc Pollefeys
IDIvan Dokmanić
Ivan Dokmanić
We propose a method for sensor array self-localization using a set of sources at unknown locations. The sources produce signals whose times of arrival are registered at the sensors. We look at the general case where neither the emission times of the sources nor the reference time frames of the receivers are known. Unlike previous work, our method directly recovers the array geometry, instead of first estimating the timing information. The key component is a new loss function which is insensitive to the unknown timings. We cast the problem as a minimization of a non-convex functional of the Euclidean distance matrix of microphones and sources subject to certain non-convex constraints. After convexification, we obtain a semidefinite relaxation which gives an approximate solution; subsequent refinement on the proposed loss via the Levenberg-Marquardt scheme gives the final locations. Our method achieves state-of-the-art performance in terms of reconstruction accuracy, speed, and the ability to work with a small number of sources and receivers. It can also handle missing measurements and exploit prior geometric and temporal knowledge, for example if either the receiver offsets or the emission times are known, or if the array contains compact subarrays with known geometry.