Optimal algorithm found for anytime regret with two experts.
problem Minimizing regret in prediction with two experts when time horizon is unknown.
method Designing a minimax optimal algorithm using ideas from stochastic calculus.
result Proved the optimal regret is γ√t / 2 for all time steps t.
We introduce a new sequential Monte Carlo algorithm we call the particle cascade. The particle cascade is an asynchronous, anytime alternative to traditional particle filtering algorithms. It uses no barrier synchronizations which leads to improved particle throughput and memory efficiency. It is an anytime algorithm i…
Enhances early-exit neural networks for anytime classification.
problem Lack of guaranteed prediction quality improvement with longer computation time.
method Post-hoc modification based on Product-of-Experts to enforce conditional monotonicity.
result Achieves conditional monotonicity in prediction quality, enabling anytime classification.
I introduce and analyse an anytime version of the Optimally Confident UCB (OCUCB) algorithm designed for minimising the cumulative regret in finite-armed stochastic bandits with subgaussian noise. The new algorithm is simple, intuitive (in hindsight) and comes with the strongest finite-time regret guarantees for a hori…
New algorithm guarantees performance on noisy data.
problem Learning with noisy data and heavy-tailed distributions.
method Anytime online-to-batch conversion for smooth objectives.
result Stochastic gradient-based algorithm with sub-Gaussian error bounds.
Optimizes random forest inference by defining step order to maximize accuracy.
problem Limited inference time in resource-constrained systems.
method Designs anytime random forest algorithm on step granularity, proposing optimal step order.
result Backward Squirrel Order performs nearly as well as the optimal step order.
APGAI identifies good arms anytime with fixed budget.
problem Identifying a good arm with a fixed sampling budget.
method An anytime algorithm for good arm identification in stochastic bandits.
result APGAI achieves efficient detection of good arms with upper bounds on probability of error and sampling complexity.
We introduce a novel incremental decision tree learning algorithm, Hoeffding Anytime Tree, that is statistically more efficient than the current state-of-the-art, Hoeffding Tree. We demonstrate that an implementation of Hoeffding Anytime Tree---"Extremely Fast Decision Tree", a minor modification to the MOA implementat…
Paper proposes an online transfer learning framework using online bagging.
problem Difficulty in obtaining sufficient labeled data in the target domain.
method Ensemble approach with online bagging for anytime transfer learning.
result Effectiveness of the proposed algorithms demonstrated on real data sets.
New algorithm identifies best arm with optimal budget usage.
problem Identifying the arm with the highest mean reward from multiple options.
method Proposes Almost Tracking, a closed-form algorithm for anytime best arm identification.
result Proven to be rate-optimal and outperforms existing algorithms.
An online reinforcement learning algorithm is anytime if it does not need to know in advance the horizon T of the experiment. A well-known technique to obtain an anytime algorithm from any non-anytime algorithm is the "Doubling Trick". In the context of adversarial or stochastic multi-armed bandits, the performance of …
This work accelerates gradient descent with anytime convergence guarantees.
problem Improving the convergence rate of gradient descent methods.
method Proposes a stepsize schedule for gradient descent that achieves anytime convergence rates.
result Gradient descent can achieve convergence rates of O(T−1.119) for any stopping time T. Optimizes SGD for anytime neural networks, improving accuracy.
problem Training networks that produce increasingly accurate outputs over time.
method Orthogonalized SGD optimizer for nested architectures.
result Significantly improves generalization accuracy of anytime networks.
Belief Propagation has been widely used for marginal inference, however it is slow on problems with large-domain variables and high-order factors. Previous work provides useful approximations to facilitate inference on such models, but lacks important anytime properties such as: 1) providing accurate and consistent mar…
GAAVI offers anytime-valid tests for CMF global null and contrasts.
problem Inference on the conditional mean function for high confidence decisions.
method Asymptotic anytime-valid tests for CMF global null and contrasts.
result Achieves asymptotic type-I error guarantees, power one, and optimal sample complexity.
This paper presents a new anytime algorithm for the marginal MAP problem in graphical models. The algorithm is described in detail, its complexity and convergence rate are studied, and relations to previous theoretical results for the problem are discussed. It is shown that the algorithm runs in polynomial-time if the …
The study introduces anytime learning schedules for large language models without fixed horizons.
problem Training large language models without knowing the total training horizon.
method Theoretical analysis and weight averaging to create anytime learning schedules.
result Theoretical and empirical evidence shows that weight averaging with simple step sizes can achieve comparable final loss to well-tuned cosine schedules.
CSA fills a gap in RLVR-trained LLM deployment by providing anytime-valid selective risk control.
problem Deployment of RLVR-trained LLMs in regulated organizations requires a safety certificate for every round without waiting for long-run averages.
method CSA uses a (test statistic, validity guarantee, deployment rule) framework to fill the gap, maintaining a Ville-type e-process per threshold on a Bonferroni grid.
result CSA provides the first anytime-valid selective risk control for RLVR-trained LLMs, matching the long-run average certification rate and satisfying pathwise validity and non-refusing deployment on every cell.
Transforms any test into anytime-valid with sample savings.
problem Sequential data invalidates classical test guarantees.
method Predicts test outcomes to create anytime-valid stopping rules.
result Ensures Type-I error control and near-optimal power.
CITE algorithm provides anytime-valid certification of model outputs.
problem Challenges in controlling error levels in LLM self-consistency.
method Certification by Intersection-union Testing with E-processes (CITE) algorithm.
result Provable control of false certification at any prescribed level under arbitrary stopping rules.
We propose a new anytime hierarchical clustering method that iteratively transforms an arbitrary initial hierarchy on the configuration of measurements along a sequence of trees we prove for a fixed data set must terminate in a chain of nested partitions that satisfies a natural homogeneity requirement. Each recursive …
The pioneer deep neural networks (DNNs) have emerged to be deeper or wider for improving their accuracy in various applications of artificial intelligence. However, DNNs are often too heavy to deploy in practice, and it is often required to control their architectures dynamically given computing resource budget, i.e., …
In this paper, we study the behavior of the Hedge algorithm in the online stochastic setting. We prove that anytime Hedge with decreasing learning rate, which is one of the simplest algorithm for the problem of prediction with expert advice, is surprisingly both worst-case optimal and adaptive to the easier stochastic …
Unified framework for ensemble sampling in nonlinear contextual bandits with provable regret bounds.
problem Efficient exploration in nonlinear contextual bandits with unknown feature dimensions.
method Developed GLM-ES and Neural-ES for generalized linear and neural contextual bandits, respectively, using maximum likelihood estimation on randomly perturbed data.
result Unified high-probability frequentist regret bounds for GLM-ES and Neural-ES, matching state-of-the-art results.
Paper studies continuous prediction with experts' advice using differential equations.
problem Continuous prediction with experts' advice in online learning.
method Continuous-time stochastic calculus and differential equations.
result Improved guarantees for quantile regret with continuous-time algorithm.
Monte Carlo algorithms simulate some prescribed number of samples, taking some random real time to complete the computations necessary. This work considers the converse: to impose a real-time budget on the computation, which results in the number of samples simulated being random. To complicate matters, the real time t…
Box Thirding identifies the best arm efficiently under limited samples.
problem Efficiently identifying the best arm with limited sampling.
method Iterative ternary comparison of arms, discarding the weakest and exploring the best.
result Achieves comparable performance to Successive Halving with less predefined parameters.
This paper studies the deviations of the regret in a stochastic multi-armed bandit problem. When the total number of plays n is known beforehand by the agent, Audibert et al. (2009) exhibit a policy such that with probability at least 1-1/n, the regret of the policy is of order log(n). They have also shown that such a …
Efficient algorithms find optimal monotone transforms for calibration under strictly convex losses.
problem Calibrating estimations to improve performance with monotone transforms.
method Proposed linear-time and space algorithm for finding optimal monotone transforms for specific loss functions. Also proposed an anytime algorithm with linear space and pseudo-linearithmic time complexity.
result Optimal monotone transforms are unique and can be found efficiently for various strictly convex loss functions.
Paper tackles safe combinatorial semi-bandits with risk constraints.
problem Safe combinatorial semi-bandits with risk constraints.
method Formulated probably anytime-safe constraint, designed PASCombUCB algorithm.
result PASCombUCB is almost asymptotically optimal in minimizing regret.
Paper develops a new watermarking framework for LLMs.
problem Efficiently distinguishing machine-generated from human text.
method Develops Anchored E-Watermarking framework for anytime-valid inference.
result Framework enables valid, anytime-inference with reduced sample efficiency.
We propose an adaptive sampling approach for multiple testing which aims to maximize statistical power while ensuring anytime false discovery control. We consider n distributions whose means are partitioned by whether they are below or equal to a baseline (nulls), versus above the baseline (actual positives). In addi…
Paper offers anytime-valid inference for causal parameters using DML.
problem Classic DML is only valid asymptotically for a fixed sample size.
method Time-uniform DML results for anytime-valid inference.
result Valid inference at any arbitrary stopping time.
Anytime MiniBatch speeds up online distributed optimization by handling slow nodes.
problem Mitigating the impact of slow nodes (stragglers) in distributed optimization.
method Proposes an online distributed optimization method that averages minibatch gradients via consensus rounds.
result Prevents stragglers from slowing progress without wasting work.
Adaptive auditing improves AI robustness testing with anytime-valid guarantees.
problem Cost and time of annotation limit rigorous AI failure mode characterization.
method Introduces hypothesis testing framework for adaptive audits using SAVI.
result Proves anytime-valid type-I error control and robustness certification.
We consider the thresholding bandit problem, whose goal is to find arms of mean rewards above a given threshold θ, with a fixed budget of T trials. We introduce LSA, a new, simple and anytime algorithm that aims to minimize the aggregate regret (or the expected number of mis-classified arms). We prove that our algo…
We describe an embarrassingly parallel, anytime Monte Carlo method for likelihood-free models. The algorithm starts with the view that the stochasticity of the pseudo-samples generated by the simulator can be controlled externally by a vector of random numbers u, in such a way that the outcome, knowing u, is determinis…
We introduce an approximate search algorithm for fast maximum a posteriori probability estimation in probabilistic programs, which we call Bayesian ascent Monte Carlo (BaMC). Probabilistic programs represent probabilistic models with varying number of mutually dependent finite, countable, and continuous random variable…
Develops an anytime-valid framework for optimal policy identification from logged contextual bandit data.
problem Selecting the optimal policy from a candidate policy class while monitoring evidence continuously.
method Constructs a time-indexed set that retains the true optimal policy set uniformly over time.
result The procedure allows the analyst to monitor policy values, eliminate clearly suboptimal policies, and stop at data-dependent times without invalidating inference.
New findings show many popular bandit algorithms are unstable, contradicting minimax optimality.
problem Challenges in statistical inference from bandit algorithms due to adaptive, non-i.i.d. nature.
method Analysis of stability properties of optimism-based bandit algorithms.
result Widely used minimax-optimal UCB-style algorithms are unstable.
New methods for anytime-valid off-policy inference in contextual bandits.
problem Estimating properties of hypothetical policies in adaptive experiments.
method Modern martingale techniques for comprehensive OPE inference.
result Valid anytime inference for off-policy mean reward values and entire reward distributions.
Develops anytime-valid stopping rules for SGD based on observed trajectory.
problem Stopping stochastic gradient descent (SGD) based on observed trajectory.
method Develops anytime-valid confidence sequences for stochastic gradient methods.
result Statistically valid, time-uniform stopping rules for SGD across convex and nonconvex settings.
Develops anytime-valid conformal and PAC prediction for streaming data.
problem Lack of guarantee in traditional conformal methods for sequential settings.
method Extends conformal and PAC prediction frameworks to handle streaming data.
result Provides anytime-valid prediction sets for sequential settings.
ALEXP improves model selection in linear bandits with exponential regret improvement.
problem Model selection in linear bandits is challenging due to balancing exploration and exploitation.
method ALEXP uses online learning with favorable bias-variance trade-off to emulate full-information feedback.
result ALEXP achieves an exponentially improved (logM) regret dependence on the number of models M. Four geometries govern sequential and distribution-free inference.
problem Sequential and distribution-free inference challenges.
method Four distinct admissibility geometries.
result Four classes of admissible procedures are pairwise non-nested.
We consider the problem of learning a loss function which, when minimized over a training dataset, yields a model that approximately minimizes a validation error metric. Though learning an optimal loss function is NP-hard, we present an anytime algorithm that is asymptotically optimal in the worst case, and is provably…
We present a new anytime algorithm that achieves near-optimal regret for any instance of finite stochastic partial monitoring. In particular, the new algorithm achieves the minimax regret, within logarithmic factors, for both "easy" and "hard" problems. For easy problems, it additionally achieves logarithmic individual…
Paper optimizes Bayesian optimization for complex functions with macro-actions.
problem Optimizing complex, highly uncertain functions efficiently.
method Generalized GP upper confidence bound with macro-actions for scalable lookahead.
result Asymptotically optimal anytime variant of epsilon-Macro-GPO policy.