arXiv research

Cataclysm deformations study Anosov representations and their convergence.

Survey of recent Kleinian representation convergence results.

Cataclysm deformations study Anosov representations, leading to new formulas and non-open sets.

CKA with Gaussian RBF kernels converges linearly as bandwidth increases.

We prove that a sequence of quasi-Fuchsian representations for which the critical exponent converges to the topological dimension of the boundary of the group (larger than 2), converges up to subsequence and conjugacy to a totally geodesic representation.

Temporal-difference and Q-learning learn feature representations that converge to optimal ones.

Regularized LAEs learn principal components efficiently.

Comparing different neural network representations and determining how representations evolve over time remain challenging open questions in our understanding of the function of neural networks. Comparing representations in neural networks is fundamentally difficult as the structure of representations varies greatly, e…

Random harmonic maps into spheres converge to a specific metric under strong convergence of representations.

In this paper we investigate the Hausdorff dimension of limit sets of Anosov representations. In this context we revisit and extend the framework of hyperconvex representations and establish a convergence property for them, analogue to a differentiability property. As an application of this convergence, we prove that t…

Many fractional processes can be represented as an integral over a family of Ornstein-Uhlenbeck processes. This representation naturally lends itself to numerical discretizations, which are shown in this paper to have strong convergence rates of arbitrarily high polynomial order. This explains the potential, but also s…

Paper finds sparse representation of functions using inverse scale space flow.

Bézier-GAN optimizes airfoil design by reducing shape complexity.

Many loss functions in representation learning are invariant under a continuous symmetry transformation. For example, the loss function of word embeddings (Mikolov et al., 2013) remains unchanged if we simultaneously rotate all word and context embedding vectors. We show that representation learning models for time ser…

Uniformly random permutations converge to regular representation on surface groups.

Early alignment in neural networks leads to sparse representations but hinders convergence.

We develop a stochastic target representation for Ricci flow and normalized Ricci flow on smooth, compact surfaces, analogous to Soner and Touzi's representation of mean curvature flow. We prove a verification/uniqueness theorem, and then consider geometric consequences of this stochastic representation. Based on this …

New theory explains how self-supervised learning converges, advancing AI research.

In this paper, we obtain stability results for martingale representations in a very general framework. More specifically, we consider a sequence of martingales each adapted to its own filtration, and a sequence of random variables measurable with respect to those filtrations. We assume that the terminal values of the m…

RO-TD learns sparse value functions efficiently.

Shallow neural networks can represent polynomials efficiently.

We extend contrastive learning theory for multiway classification and prove convergence guarantees.

Neural networks learn spectral representations for group composition.

RVI accelerates encoderless VI for faster convergence.

We use quantum invariants to define an analytic family of representations for the mapping class group of a punctured surface. The representations depend on a complex number A with |A| <= 1 and act on an infinite-dimensional Hilbert space. They are unitary when A is real or imaginary, bounded when |A|<1, and only densel…

EBM reduces dimensionality for estimating heterogeneous CATEs.

This paper analyzes neural networks for solving complex optimization problems.

For each oriented surface $Σ$ of genus $g$ we study a limit of quantum representations of the mapping class group arising in TQFT derived from the Kauffman bracket. We determine that these representations converge in the Fell topology to the representation of the mapping class group on $\boH(Σ)$, the space of regular f…

This paper studies nonlinear representation learning dynamics beyond the NTK regime.

We construct a sequence of primitive-stable representations of free groups into PSL(2,C) whose ranks go to infinity, but whose images are discrete with quotient manifolds that converge geometrically to a knot complement. In particular this implies that the rank and geometry of the image of a primitive-stable representa…

KL annealing helps VAEs avoid posterior collapse and overfitting.

Anderson and Canary have shown that if the algebraic limit of a sequence of discrete, faithful representations of a finitely generated group into PSL(2,C) does not contain parabolics, then it is also the sequence's geometric limit. We construct examples that demonstrate the failure of this theorem for certain sequences…

Paper shows linear convergence of ISTA and FISTA for ill-conditioned images.

This belongs to a series of papers devoted to the study of the cohomology of classifying spaces of Lie groupoids. Our aim here is to introduce and study the notion of representation up to homotopy of Lie groupoids, the resulting derived category, and to show that the adjoint representation is well defined as a represen…

Adapting functional gradients improves FGD's practicality and theoretical guarantees.

New meta-learning method handles nonlinear tasks for faster convergence.

Most model-free reinforcement learning methods leverage state representations (embeddings) for generalization, but either ignore structure in the space of actions or assume the structure is provided a priori. We show how a policy can be decomposed into a component that acts in a low-dimensional space of action represen…

Tensor networks improve integration accuracy for high-dimensional problems.

In this paper, we propose a dynamical systems perspective of the Expectation-Maximization (EM) algorithm. More precisely, we can analyze the EM algorithm as a nonlinear state-space dynamical system. The EM algorithm is widely adopted for data clustering and density estimation in statistics, control systems, and machine…

Sparsity inducing regularization is an important part for learning over-complete visual representations. Despite the popularity of $\ell_1$ regularization, in this paper, we investigate the usage of non-convex regularizations in this problem. Our contribution consists of three parts. First, we propose the leaky capped …

Paper develops a method for causal representation learning from irregular tensors.

Wide hypernetworks don't guarantee convergence under gradient descent.

In this paper, we establish Basmajian's identity for $(1,1,2)$-hyperconvex Anosov representations from a free group into $PGL(n, R)$. We then study our series identities on holomorphic families of Cantor non-conformal repellers associated to complex $(1,1,2)$-hyperconvex Anosov representations. We show that the series …

PFedRL-Rep learns shared and personalized policies for heterogeneous environments.

Multiview representation learning is very popular for latent factor analysis. It naturally arises in many data analysis, machine learning, and information retrieval applications to model dependent structures among multiple data sources. For computational convenience, existing approaches usually formulate the multiview …

The paper proves a criterion for L-space knots and their representations.

We investigate the rigidity and asymptotic properties of quantum SU(2) representations of mapping class groups. In the spherical braid group case the trivial representation is not isolated in the family of quantum SU(2) representations. In particular, they may be used to give an explicit check that spherical braid grou…

For $N \geq 2$, we study a certain sequence $(ρ_p^{(c_p)})$ of N-dimensional representations of the mapping class group of the one-holed torus arising from SO(3)-TQFT, and show that the conjecture of Andersen, Masbaum, and Ueno \cite{1} holds for these representations. This is done by proving that, in a certain basis a…