arXiv research

FSRL balances fairness and sufficiency in learning representations.

MSRL learns a representation maximizing mutual info with response variables.

Study shows how information loss and operation loss are related in feature representations.

Unified framework for fair representation learning in machine learning.

Study on how optimal representations emerge during deep learning training, focusing on the role of implicit regularization.

Enhances source domain knowledge with target data for transfer learning.

TCRI improves domain generalization by enforcing conditional independence constraints.

Unpaired multi-domain causal representation learning is possible with sufficient conditions.

Greg McShane introduced a remarkable identity for lengths of simple closed geodesics on the once punctured torus with a complete, finite volume hyperbolic structure. Bowditch later generalized this and gave sufficient conditions for the identity to hold for general type-preserving representations of a free group on two…

Many learning algorithms require categorical data to be transformed into real vectors before it can be used as input. Often, categorical variables are encoded as one-hot (or dummy) vectors. However, this mode of representation can be wasteful since it adds many low-signal regressors, especially when the number of uniqu…

Study feature representations induced by dependence between variables.

The paper explores symmetric representations of links and conditions for amphichirality.

New statistical theory explains contrastive learning effectiveness.

Proposes a deep learning method for effective data representation.

In the article the necessary and sufficient conditions for a representation of Lipschitz function of two variables as a difference of two convex functions are formulated. An algorithm of this representation is given. The outcome of this algorithm is a sequence of pairs of convex functions that converge uniformly to a p…

Intelligent agents should be able to learn useful representations by observing changes in their environment. We model such observations as pairs of non-i.i.d. images sharing at least one of the underlying factors of variation. First, we theoretically show that only knowing how many factors have changed, but not which o…

In this paper we establish necessary and sufficient conditions for the limit set of a projective Anosov representation to be a differentiable submanifold of projective space with Holder continuous derivatives. We also calculate the optimal value of the Holder constant in terms of the eigenvalue data of the Anosov repre…

We consider a general regularised interpolation problem for learning a parameter vector from data. The well known representer theorem says that under certain conditions on the regulariser there exists a solution in the linear span of the data points. This is at the core of kernel methods in machine learning as it makes…

A method identifies domain-general features using causal graph constraints and regularization.

The paper shows that certain learned representations are identifiable in function space.

We prove that for each sufficiently complicated orientable surface $S$, there exists an infinite image linear representation $ρ$ of $π_1(S)$ such that if $γ\inπ_1(S)$ is freely homotopic to a simple closed curve on $S$, then $ρ(γ)$ has finite order. Furthermore, we prove that given a sufficiently complicated orientable…

Bézier-GAN optimizes airfoil design by reducing shape complexity.

We consider immersions admitting uniform graph representations over the affine tangent space over a ball of fixed radius r>0. We show that for sufficiently small C^0-norm of the graph functions, each graph function is smooth with small C^1-norm.

In this paper, we give a necessarly and sufficient condition for orbits of linear isotropy representations of Riemannian symmetric spaces are biharmonic submanifolds in hyperspheres in Euclidean spaces. In particular, we obtain examples of biharmonic submanifolds in hyperspheres whose co-dimension is greater than one.

Develops new methods for isospectral orbifolds and regulator quotients.

Fairness has become a central issue for our research community as classification algorithms are adopted in societally critical domains such as recidivism prediction and loan approval. In this work, we consider the potential bias based on protected attributes (e.g., race and gender), and tackle this problem by learning …

We give sufficient conditions for a parametrised family of probability measures on a Riemannian manifold with boundary to be represented by random maps of class $C^k$. The conditions allow for the probability densities to approach zero towards the boundary of the manifold. We also formulate two obstructions to regular …

Let $S$ be a surface of genus $g$ at least $2$. A representation $ρ:π_1S\longrightarrow \text{PSL}_2\Bbb R$ is said to be purely hyperbolic if its image consists only of hyperbolic elements other than the identity. We may wonder under which conditions such representations arise as holonomy of a hyperbolic cone-structur…

Poly-view contrastive learning improves image representation learning.

Let $Γ$ be the fundamental group of the exterior of a knot in the three-sphere. We study deformations of representations of $Γ$ into $\mathrm{SL}_n(\mathbf{C})$ which are the sum of two irreducible representations. For such representations we give a necessary condition, in terms of the twisted Alexander polynomial, for…

the main theorem gives a sufficient condition for a n elements of SL(2,R) to generate a free group.The idea behind it is to use a nonorientable version of the Dehn-Wolpert-Goldman twist and to sew it with the original representation of a free group to get representation of the closed surfase group and then to apply Gol…

New representations of surface groups into higher-dimensional PSL generalize pleated surfaces.

Due to the ability of deep neural nets to learn rich representations, recent advances in unsupervised domain adaptation have focused on learning domain-invariant features that achieve a small error on the source domain. The hope is that the learnt representation, together with the hypothesis learnt from the source doma…

New method combines domain changes and sparse mixing for better latent variable learning.

Modern deep learning methods provide effective means to learn good representations. However, is a good representation itself sufficient for sample efficient reinforcement learning? This question has largely been studied only with respect to (worst-case) approximation error, in the more classical approximate dynamic pro…

Paper develops a theory explaining contrastive pre-training for multimodal AI.

Pantypes improve prototypical models by capturing diverse input distributions.

We prove a uniqueness result for finite-dimensional representations of the Kauffman skein algebra $\mathcal{S}_A(S)$ of a surface $S$, when $A$ is a root of unity and when the surface $S$ is a sphere with at most four punctures or a torus with at most one puncture. We show that, if two irreducible representations of $\…

We present a framework for autonomously learning a portable representation that describes a collection of low-level continuous environments. We show that these abstract representations can be learned in a task-independent egocentric space specific to the agent that, when grounded with problem-specific information, are …

We study spin structures on affine Kac-Moody symmetric spaces and obtain sufficient conditions for their existence.\ As a by product of this, we obtain a spin-c representation of certain Kac-Moody quadratic subgroups of type E.

Given an overcomplete dictionary $A$ and a signal $b$ that is a linear combination of a few linearly independent columns of $A$, classical sparse recovery theory deals with the problem of recovering the unique sparse representation $x$ such that $b = A x$. It is known that under certain conditions on $A$, $x$ can be re…

We study isometric actions of tree automorphism groups on the infinite-dimensional hyperbolic spaces. On the one hand, we exhibit a general one-parameter family of such representations and analyse the corresponding equivariant embeddings of the trees, showing that they are convex-cocompact and asymptotically isometric.…

Solves Christoffel problem for disk area measures on spheres.

LLMs' explanations are often insufficient and vary with input distribution.

This article studies quadratic semimartingale BSDEs arising in power utility maximization when the market price of risk is of BMO type. In a Brownian setting we provide a necessary and sufficient condition for the existence of a solution but show that uniqueness fails to hold in the sense that there exists a continuum …

Study character varieties for 3-punctured sphere group representations in PU(2,1).

The paper studies conjugating complex representations into real ones.

The paper characterizes isomorphic covers of surfaces and applies it to distinguish representations.