31 accepted papers
A fundamental challenge in the theory of deep learning is to understand whether gradient-based training can promote parameters belonging to certain lower-dimensional structures (e.g., sparse or low-rank sets), leading to so-called implicit bias. As a stepping stone, motivated by the proof structure…
Despite recent algorithmic advances, we still lack principled ways to leverage the well-documented rescaling symmetries in ReLU neural network parameters. While two properly rescaled weights implement the same function, the training dynamics can be dramatically different. To offer a fresh perspectiv…
Robustness with respect to weight perturbations underpins guarantees for generalization, pruning and quantization. Existing guarantees rely on *Lipschitz bounds in parameter space*, cover only plain feed-forward MLPs, and break under the ubiquitous neuron-wise rescaling symmetry of ReLU networks. We…
While conservation laws in gradient flow training dynamics are well understood for (mostly shallow) ReLU and linear networks, their study remains largely unexplored for more practical architectures. For this, we first show that basic building blocks such as ReLU (or linear) shallow networks, with or…
This work introduces the first toolkit around path-norms that fully encompasses general DAG ReLU networks with biases, skip connections and any operation based on the extraction of order statistics: max pooling, GroupSort etc. This toolkit notably allows us to establish generalization bounds for mod…
Conservation laws are well-established in the context of Euclidean gradient flow dynamics, notably for linear or ReLU neural network training. Yet, their existence and principles for non-Euclidean geometries and momentum-based dynamics remain largely unknown. In this paper, we characterize "all" con…
Understanding the geometric properties of gradient descent dynamics is a key ingredient in deciphering the recent success of very large machine learning models. A striking observation is that trained over-parameterized models retain some properties of the optimization initialization. This "implicit…
Given a training set, a loss function, and a neural network architecture, it is often taken for granted that optimal network parameters exist, and a common practice is to apply available optimization algorithms to search for them. In this work, we show that the existence of an optimal solution is no…
This work studies the estimation of many statistical quantiles under differential privacy. More precisely, given a distribution and access to i.i.d. samples from it, we study the estimation of the inverse of its cumulative distribution function (the quantile function) at specific points. For instanc…
We introduce a regularization loss based on kernel mean embeddings with rotation-invariant kernels on the hypersphere (also known as dot-product kernels) for self-supervised learning of image representations. Besides being fully competitive with the state of the art, our method significantly reduces…
Approximating a matrix by a product of few sparse factors whose supports possess the butterfly structure, which is common to many fast transforms, is key to learn fast transforms and speed up algorithms for inverse problems. We introduce a hierarchical approach that recursively factorizes the consid…
Diffusing a graph signal at multiple scales requires to compute the action of the exponential of as many versions of the Laplacian matrix. Considering the truncated Chebyshev polynomial approximation of the exponential, we derive a tightened bound on the approximation error, allowing thus for a bett…
Matrix factorization with sparsity constraints plays an important role in many machine learning and signal processing problems such as dictionary learning, data visualization, dimension reduction. Among the most popular tools for sparse matrix factorization are proximal algorithms, a family of algor…
We tackle the problem of producing compact models, maximizing their accuracy for a given model size. A standard solution is to train networks with Quantization Aware Training, where the weights are quantized during training and the gradients approximated with the Straight-Through Estimator. In this…
In this paper, we address the problem of reducing the memory footprint of convolutional network architectures. We introduce a vector quantization method that aims at preserving the quality of the reconstruction of the network outputs rather than its weights. The principle of our approach is that it…
Acoustic echoes retrieval is a research topic that is gaining importance in many speech and audio signal processing applications such as speech enhancement, source separation, dereverberation and room geometry estimation. This work proposes a novel approach to blindly retrieve the off-grid timing of…
We propose a numerical interferometry method for identification of optical multiply-scattering systems when only intensity can be measured. Our method simplifies the calibration of optical transmission matrices from a quadratic to a linear inverse problem by first recovering the phase of the measure…
This work addresses the problem of learning from large collections of data with privacy guarantees. The sketched learning framework proposes to deal with the large scale of datasets by compressing them into a single vector of generalized random moments, from which the learning task is then performed…
Modern neural networks are over-parametrized. In particular, each rectified linear hidden unit can be modified by a multiplicative factor by adjusting input and out- put weights, without changing the rest of the network. Inspired by the Sinkhorn-Knopp algorithm, we introduce a fast iterative method…
In this work, we present new theoretical results on sparse recovery guarantees for a greedy algorithm, orthogonal matching pursuit (OMP), in the context of continuous parametric dictionaries, i.e., made up of an infinite uncountable number of atoms. We build up a family of dictionaries for which k-s…
This work features a new algorithm, CASCADE, which leverages a structured cosparse prior across channels to address the multichannel audio declipping problem. CASCADE technique outperforms the state-of-the-art method A-SPADE applied on each channel separately in all tested settings, while retaining…
Accelerating the solution of the Lasso problem becomes crucial when scaling to very high dimensional data. In this paper, we propose a way to combine two existing acceleration techniques: safe screening tests, which simplify the problem by eliminating useless dictionary atoms; and the use of structu…
In this paper, we address the problem of high-dimensional k-means clustering in a large-scale setting, i.e. for datasets that comprise a large number of items. Sketching techniques have already been used to deal with this “large-scale” issue, by compressing the whole dataset into a single vector of…
To work at scale, a complete image indexing system comprises two components: An inverted file index to restrict the actual search to only a subset that should contain most of the items relevant to the query; An approximate distance computation mechanism to rapidly scan these lists. While supervised…
The Lloyd-Max algorithm is a classical approach to perform K-means clustering. Unfortunately, its cost becomes prohibitive as the training dataset grows large. We propose a compressive version of K-means (CKM), that estimates cluster centers from a sketch, i.e. from a drastically compressed represen…
We build upon recent advances in graph signal processing to propose a faster spectral clustering algorithm. Indeed, classical spectral clustering is based on the computation of the first k eigenvectors of the similarity matrix' Laplacian, whose computation cost, even for sparse matrices, becomes pro…
Signal processing on graphs is a recent research domain that seeks to extend classical signal processing tools such as the Fourier transform to irregular domains given by a graph. In such a graph setting, a way to rapidly apply the Fourier transform, i.e. a Fast Fourier Transform (FFT), is lacking.…
Indoor acoustic source localization can be efficiently performed by modeling the sound propagation in the room, and by solving the arising inverse problem by means of cosparse regularization and convex optimization techniques. However, previous methods relying on this approach used to assume the kno…
This paper investigates the problem of estimating the shape or the boundary impedance of a vibrating membrane from acoustic measurements in a limited sub-domain of the membrane. In acoustics, polygonal room shapes are usually estimated through room impulse response measurements. Impedance values of…
Learning parameters from voluminous data can be prohibitive in terms of memory and computational requirements. We propose a "compressive learning" framework where we first sketch the data by computing random generalized moments of the underlying probability distribution, then estimate mixture model…
Dictionary learning aims at finding a frame (called dictionary) in which some training data admits a sparse representation. Traditional dictionary learning is limited to relatively small-scale problems, because high-dimensional dense dictionaries can be costly to manipulate, both at the learning sta…