arXiv research

First sample-efficient algorithm for learning EFCE in bandit feedback settings.

Dirac-harmonic maps are uncoupled under certain conditions.

Study uncoupled solutions to Dirac-Yang-Mills equations on spin manifolds.

Uncoupled regression is the problem to learn a model from unlabeled data and the set of target values while the correspondence between them is unknown. Such a situation arises in predicting anonymized targets that involve sensitive information, e.g., one's annual income. Since existing methods for uncoupled regression …

New algorithms converge faster to Nash equilibrium in zero-sum games with bandit feedback.

Paper establishes identifiability and achievability for causal representation learning.

No-regret optimization for time-varying functions using uncertainty injection.

No-regret learning with strategic experts, incentivized.

Efficient algorithm converges to Nash equilibrium in bilinear problems with bandit feedback.

No-regret learning fails to converge to Nash equilibria in mixed strategies.

New method learns functions without paired data using mediating variables.

Isotonic regression is a standard problem in shape-constrained estimation where the goal is to estimate an unknown nondecreasing regression function $f$ from independent pairs $(x_i, y_i)$ where $\mathbb{E}[y_i]=f(x_i), i=1, \ldots n$. While this problem is well understood both statistically and computationally, much l…

We study Dirac-harmonic maps from surfaces to manifolds with torsion, which is motivated from the superstring action considered in theoretical physics. We discuss analytic and geometric properties of such maps and outline an existence result for uncoupled solutions.

Multi-domain translation seeks to learn a probabilistic coupling between marginal distributions that reflects the correspondence between different domains. We assume that data from different domains are generated from a shared latent representation based on a structural equation model. Under this assumption, we show th…

New insights link no-regret learning to online conformal prediction in adversarial settings.

Paper proposes no-regret algorithms for private GP bandit optimization.

New framework uses PDE for no-regret generative modeling.

We prove existence results for Dirac-harmonic maps using index theoretical tools. They are mainly interesting if the source manifold has dimension 1 or 2 modulo 8. Our solutions are uncoupled in the sense that the underlying map between the source and target manifolds is a harmonic map.

Paper explores rate-preserving reductions between Blackwell approachability and no-regret learning.

Kernel-based function approximation improves reinforcement learning performance.

New research shows no-regret learning is impossible in Markov games under certain assumptions.

Node2Grids uncouples GCN training for large graphs, saving memory and computation.

Sequential prediction problems such as imitation learning, where future observations depend on previous predictions (actions), violate the common i.i.d. assumptions made in statistical learning. This leads to poor performance in theory and often in practice. Some recent approaches provide stronger guarantees in this se…

The paper tackles cooperative RL with function approximation, achieving near-optimal learning with limited communication.

A new model for simulating cloth manipulation in robots, accurate to within 1cm.

Paper analyzes GP-EI for Bayesian optimization with no regret and provides guidance on choosing incumbents.

New algorithms achieve no-regret learning even with adversarial transitions and losses.

Paper proposes OPF policy for fair resource allocation with sublinear regret.

DORIS algorithm achieves no-regret learning in Markov games with adversarial opponents.

A method for safe online classification reduces test costs while maintaining low error rates.

We classify quasilinear systems in Riemann invariants whose characteristic webs are linearizable on every solution. Although the linearizability of an individual web is a rather nontrivial differential constraint, the requirement of linearizability of characteristic webs on all solutions imposes simple second-order con…

R2-B2 optimizes game interactions with recursive reasoning.

OMWU shows last iterate convergence in convex-concave games.

Paper establishes no-regret property for practical EGO optimization.

A new mechanism reduces expert belief regret in online forecasting.

Paper proves existence of Dirac-harmonic maps with trivial index.

We consider the use of no-regret algorithms to compute equilibria for particular classes of convex-concave games. While standard regret bounds would lead to convergence rates on the order of $O(T^{-1/2})$, recent work \citep{RS13,SALS15} has established $O(1/T)$ rates by taking advantage of a particular class of optimi…

Counterfactual Regret Minimization (CFR) has found success in settings like poker which have both terminal states and perfect recall. We seek to understand how to relax these requirements. As a first step, we introduce a simple algorithm, local no-regret learning (LONR), which uses a Q-learning-like update rule to allo…

This paper tackles no-regret learning for fair multi-agent social welfare optimization.

Recent work has demonstrated that problems-- particularly imitation learning and structured prediction-- where a learner's predictions influence the input-distribution it is tested on can be naturally addressed by an interactive approach and analyzed using no-regret online learning. These approaches to imitation learni…

In the present report, by using the Stokes-Helmholtz decomposition theorem the 3-dimensional Navier-Stokes equation (NSE) is uncoupled and transformed into a scalar equation for the velocity potential when the flow field is toroidal. The dynamics of the velocity potential is independent of the vector potential. The red…

The continuous-time random walk (CTRW) is a pure-jump stochastic process with several applications in physics, but also in insurance, finance and economics. A definition is given for a class of stochastic integrals driven by a CTRW, that includes the Ito and Stratonovich cases. An uncoupled CTRW with zero-mean jumps is…

This paper proposes a new portfolio allocation method using LLMs to outperform traditional strategies.

New concept of proper-calibeating extends classic calibrated forecasts to proper scoring rules.

Motivated by the sigma model limit of multicomponent Ginzburg-Landau theory, a version of the Faddeev-Skyrme model is considered in which the scalar field is coupled dynamically to a one-form field called the supercurrent. This coupled model is investigated in the general setting where physical space is an oriented Rie…

Many prediction domains, such as ad placement, recommendation, trajectory prediction, and document summarization, require predicting a set or list of options. Such lists are often evaluated using submodular reward functions that measure both quality and diversity. We propose a simple, efficient, and provably near-optim…

We consider a family of learning strategies for online optimization problems that evolve in continuous time and we show that they lead to no regret. From a more traditional, discrete-time viewpoint, this continuous-time approach allows us to derive the no-regret properties of a large class of discrete-time algorithms i…

The paper tackles robust policy learning from multiple data sources.