Safe Adaptive Importance Sampling

Importance sampling has become an indispensable strategy to speed up optimization algorithms for large-scale applications. Improved adaptive variants - using importance values defined by the complete gradient information which changes during optimization - enjoy favorable theoretical properties, but are typically computationally infeasible. In this paper we propose an efficient approximation of gradient-based sampling, which is based on safe bounds on the gradient. The proposed sampling distribution is (i) provably the best sampling with respect to the given bounds, (ii) always better than uniform sampling and fixed importance sampling and (iii) can efficiently be computed - in many applications at negligible extra cost. The proposed sampling scheme is generic and can easily be integrated into existing algorithms. In particular, we show that coordinate-descent (CD) and stochastic gradient descent (SGD) can enjoy significant a speed-up under the novel scheme. The proven efficiency of the proposed sampling is verified by extensive numerical testing.

Trust-region methods based on radial basis functions with application to biomedical imaging

Rodrigue Oeuvray

We have developed a new derivative-free algorithm based on Radial Basis Functions (RBFs). Derivative-free optimization is an active field of research and several algorithms have been proposed recently. Problems of this nature in the industrial setting are quite frequent. The reason is that in a number of applications the optimization process contains simulation packages which are treated as black boxes. The development of our own algorithm was originally motivated by an application in biomedical imaging: the medical image registration problem. The particular characteristics of this problem have incited us to develop a new optimization algorithm based on trust-region methods. However it has been designed to be generic and to be applied to a wide range of problems. The main originality of our approach is the use of RBFs to build the models. In particular we have adapted the existing theory based on quadratic models to our own models and developed new procedures especially designed for models based on RBFs. We have tested our algorithm called BOOSTERS against state-of-the-art methods (UOBYQA, NEWUOA, DFO). On the medical image registration problem, BOOSTERS appears to be the method of choice. The tests on problems from the CUTEr collection show that BOOSTERS is comparable to, but not better than other methods on small problems (size 2-20). It is performing very well for medium size problems (20-80). Moreover, it is able to solve problems of dimension 200, which is considered very large in derivative-free optimization. We have also developed a new class of algorithms combining the robustness of derivative-free algorithms with the faster rate of convergence characterizing Newtonlike-methods. In fact, they define a new class of algorithms lying between derivative-free optimization and quasi-Newton methods. These algorithms are built on the skeleton of our derivative-free algorithm but they can incorporate the gradient when it is available. They can be interpreted as a way of doping derivative-free algorithms with derivatives. If the derivatives are available at each iteration, then our method can be seen as an alternative to quasi-Newton methods. At the opposite, if the derivatives are never evaluated, then the algorithm is totally similar to BOOSTERS. It is a very interesting alternative to existing methods for problems whose objective function is expensive to evaluate and when the derivatives are not available. In this situation, the gradient can be approximated by finite differences and its costs corresponds to n additional function evaluations assuming that Rn is the domain of definition of the objective function. We have compared our method with CFSQP and BTRA, two gradient-based algorithms, and the results show that our doped method performs best. We have also a theoretical analysis of the medical image registration problem based on maximization of mutual information. Most of the current research in this field is concentrated on registration based on nonlinear image transformation. However, little attention has been paid to the theoretical properties of the optimization problem. In our analysis, we focus on the continuity and the differentiability of the objective function. We show in particular that performing a registration without extension of the reference image may lead to discontinuities in the objective function. But we demonstrate that, under some mild assumptions, the function is differentiable almost everywhere. Our analysis is important from an optimization point of view and conditions the choice of a solver. The usual practice is to use generic optimization packages without worrying about the differentiability of the objective function. But the use of gradient-based methods when the objective function is not differentiable may result in poor performance or even in absence of convergence. One of our objectives with this analysis is also that practitioners become aware of these problems and to propose them new algorithms having a potential interest for their applications.

EPFL2005

An Accelerated First-Order Method for Non-convex Optimization on Manifolds

Nicolas Boumal, Christopher Arnold Criscitiello

We describe the first gradient methods on Riemannian manifolds to achieve accelerated rates in the non-convex case. Under Lipschitz assumptions on the Riemannian gradient and Hessian of the cost function, these methods find approximate first-order critical points faster than regular gradient descent. A randomized version also finds approximate second-order critical points. Both the algorithms and their analyses build extensively on existing work in the Euclidean case. The basic operation consists in running the Euclidean accelerated gradient descent method (appropriately safe-guarded against non-convexity) in the current tangent space, then moving back to the manifold and repeating. This requires lifting the cost function from the manifold to the tangent space, which can be done for example through the Riemannian exponential map. For this approach to succeed, the lifted cost function (called the pullback) must retain certain Lipschitz properties. As a contribution of independent interest, we prove precise claims to that effect, with explicit constants. Those claims are affected by the Riemannian curvature of the manifold, which in turn affects the worst-case complexity bounds for our optimization algorithms.

SPRINGER2022

Sparse Gradient Optimization and its Applications in Image Processing

Nikolaos Arvanitopoulos Darginis

Millions of digital images are captured by imaging devices on a daily basis. The way imaging devices operate follows an integral process from which the information of the original scene needs to be estimated. The estimation is done by inverting the integral process of the imaging device with the use of optimization techniques. This linear inverse problem, the inversion of the integral acquisition process, is at the heart of several image processing applications such as denoising, deblurring, inpainting, and super-resolution. We describe in detail the use of linear inverse problems in these applications. We review and compare several state-of-the-art optimization algorithms that invert this integral process. Linear inverse problems are usually very difficult to solve. Therefore, additional prior assumptions need to be introduced to successfully estimate the output signal. Several priors have been suggested in the research literature, with the Total Variation (TV) being one of the most prominent. In this thesis, we review another prior, the l0 pseudo-norm over the gradient domain. This prior allows full control over how many non-zero gradients are retained to approximate prominent structures of the image. We show the superiority of the l0 gradient prior over the TV prior in recovering genuinely piece-wise constant signals. The l0 gradient prior has shown to produce state-of-the-art results in edge-preserving image smoothing. Moreover, this general prior can be applied to several other applications, such as edge extraction, clip-art JPEG artifact removal, non-photorealistic image rendering, detail magnification, and tone mapping. We review and evaluate several state-of-the-art algorithms that solve the optimization problem based on the l0 gradient prior. Subsequently we apply the l0 gradient prior to two applications where we show superior results as compared to the current state-of-the-art. The first application is that of single-image reflection removal. Existing solutions to this problem have shown limited success because of the highly ill-posed nature of the problem. We show that the standard l0 gradient prior with a modified data-fidelity term based on the Laplacian operator is able to sufficiently remove unwanted reflections from images in many realistic scenarios. We conduct extensive experiments and show that our method outperforms the state-of-the-art. In the second application of haze removal from visible-NIR image pairs we propose a novel optimization framework, where the prior term penalizes the number of non-zero gradients of the difference between the output and the NIR image. Due to the longer wavelengths of NIR, an image taken in the NIR spectrum suffers significantly less from haze artifacts. Using this prior term, we are able to transfer details from the haze-free NIR image to the final result. We show that our formulation provides state-of-the-art results compared to haze removal methods that use a single image and also to those that are based on visible-NIR image pairs.