arXiv research

Paper tackles instance-dependent label noise by approximating it with part-dependent noise.

This paper identifies and estimates the label noise transition matrix without ground truth labels.

In learning with noisy labels, for every instance, its label can randomly walk to other classes following a transition distribution which is named a noise model. Well-studied noise models are all instance-independent, namely, the transition depends only on the original label but not the instance itself, and thus they a…

We study learning latent models with multi-instance weak supervision.

New method transfers instances between domains using Gibbs Sampling and RBM.

The notion of $n$-transitivity can be carried over from groups of diffeomorphisms on a manifold $M$ to groups of bisections of a Lie groupoid over $M$. The main theorem states that the $n$-transitivity is fulfilled for all $n\in\mathbb N$ by an arbitrary group of $C^r$-bisections of a Lie groupoid $Γ$ of class $C^r$, w…

Paper proposes an alternative to anchor points for learning with noisy labels.

A method to approximate instance-dependent label noise using instance-confidence embedding.

A similarity label indicates whether two instances belong to the same class while a class label shows the class of the instance. Without class labels, a multi-class classifier could be learned from similarity-labeled pairwise data by meta classification learning. However, since the similarity label is less informative …

Adaptive algorithm identifies best arm with abstention, showing phase transition from polynomial to exponential error probability.

A new method handles label noise by leveraging causal information.

At the heart of technology transitions lie complex processes of social and industrial dynamics. The quantitative study of sustainability transitions requires modelling work, which necessitates a theory of technology substitution. Many, if not most, contemporary modelling approaches for future technology pathways overlo…

Paper establishes first instance-dependent lower bound for PAC reinforcement learning.

New model shows natural language exhibits phase transition similar to physics.

New algorithm optimally evaluates policies with linear approximations.

The paper examines stability of shares in Proof of Stake protocol, identifying different investor behaviors and phase transitions.

Reward hacking exploits misspecified rewards, affecting agent capabilities and true performance.

New algorithm reduces regret by allowing free exploration in multi-armed bandits.

A new method learns robust policies from offline data with latent structures.

Markov reward processes (MRPs) are used to model stochastic phenomena arising in operations research, control engineering, robotics, and artificial intelligence, as well as communication and transportation networks. In many of these cases, such as in the policy evaluation problem encountered in reinforcement learning, …

Unified framework for learning with indirect supervision signals.

Improves GCNNs with node transition probabilities and DropNode regularization.

We focus on the problem of search in the multilingual setting. Examining the problems of next-sentence prediction and inverse cloze, we show that at large scale, instance-based transfer learning is surprisingly effective in the multilingual setting, leading to positive transfer on all of the 35 target languages and two…

Many optimization problems can be cast into the maximum satisfiability (MAX-SAT) form, and many solvers have been developed for tackling such problems. To evaluate a MAX-SAT solver, it is convenient to generate hard MAX-SAT instances with known solutions. Here, we propose a method of generating weighted MAX-2-SAT insta…

Combinatorial auctions are formulated as frustrated lattice gases on sparse random graphs, allowing the determination of the optimal revenue by methods of statistical physics. Transitions between computationally easy and hard regimes are found and interpreted in terms of the geometric structure of the space of solution…

This paper develops the Jungle model in a credit portfolio framework. The Jungle model is able to model credit contagion, produce doubly-peaked probability distributions for the total default loss and endogenously generate quasi phase transitions, potentially leading to systemic credit events which happen unexpectedly …

Natural metrics (Sasaki metric, Cheeger-Gromoll metric, Kaluza-Klein metrics etc.. ) on the tangent bundle of a Riemannian manifold is a central topic in Riemannian geometry. Generalized Cheeger-Gromoll metrics is a family of natural metrics $h_{p,q}$ depending on two parameters with $p\in\mathbb{R}$ and $q\geq0$. This…

We study the problem of learning the transition matrices of a set of Markov chains from a single stream of observations on each chain. We assume that the Markov chains are ergodic but otherwise unknown. The learner can sample Markov chains sequentially to observe their states. The goal of the learner is to sequentially…

Optimal sequential testing for Markovian data with lower and upper bounds.

The covariance graph (aka bi-directed graph) of a probability distribution $p$ is the undirected graph $G$ where two nodes are adjacent iff their corresponding random variables are marginally dependent in $p$. In this paper, we present a graphical criterion for reading dependencies from $G$, under the assumption that $…

This work deals with the structure of the isometry group of pseudo-Riemannian 2-step nilmanifolds. We study the action by isometries of several groups and we construct examples showing substantial differences with the Riemannain situation; for instance the action of the nilradical of the isometry group does not need to…

We address the problem of estimating the mixing time of a Markov chain from a single trajectory of observations. Unlike most previous works which employed Hilbert space methods to estimate spectral gaps, we opt for an approach based on contraction with respect to total variation. Specifically, we estimate the contracti…

Method reveals dissimilarity in alloys' Curie temperatures.

Differentiable relaxation for inferring partial orders from noisy linear data.

Study phase transitions in identifying infected individuals using group testing.

A group of transition probability functions form a Shannon's channel whereas a group of truth functions form a semantic channel. Label learning is to let semantic channels match Shannon's channels and label selection is to let Shannon's channels match semantic channels. The Channel Matching (CM) algorithm is provided f…

Study improves model robustness in noisy datasets.

New insights on offline RL with state aggregation and trajectory data.

Elo rating outperforms complex models in skill estimation despite model misspecification.

An intrinsic problem of classifiers based on machine learning (ML) methods is that their learning time grows as the size and complexity of the training dataset increases. For this reason, it is important to have efficient computational methods and algorithms that can be applied on large datasets, such that it is still …

Motivated by applications in recommender systems, web search, social choice and crowdsourcing, we consider the problem of identifying the set of top $K$ items from noisy pairwise comparisons. In our setting, we are non-actively given $r$ pairwise comparisons between each pair of $n$ items, where each comparison has noi…

We investigate implicit regularization schemes for gradient descent methods applied to unpenalized least squares regression to solve the problem of reconstructing a sparse signal from an underdetermined system of linear measurements under the restricted isometry assumption. For a given parametrization yielding a non-co…

Geometric framework links clustering accuracy to structural recovery.

Because no closed timelike curve (CTC) on a Lorentzian manifold can be deformed to a point, any such manifold containing a CTC must have a topological feature, to be called a timelike wormhole, that prevents the CTC from being deformed to a point. If all wormholes have horizons, which typically seems to be the case in …

The study classifies Einstein metrics on a specific total space, revealing various behaviors and transitions.

Paper tackles anomaly detection in restless Markov arms with unknown TPMs.

Study shows it's impossible to count communities without finding them.

This paper studies the statistical theory of batch data reinforcement learning with function approximation. Consider the off-policy evaluation problem, which is to estimate the cumulative value of a new target policy from logged history generated by unknown behavioral policies. We study a regression-based fitted Q iter…