Search

Two-time-scale stochastic approximation is an iterative algorithm used in applications such as optimization, reinforcement learning, and control. Finite-time analysis of these algorithms has primarily focused on fixed point iterations where both time-scales have contractive mappings. In this paper, we study two-time-scale iterations, where the slower time-scale has a non-expansive mapping. For such algorithms, the slower time-scale can be considered a stochastic inexact Krasnoselskii-Mann iteration. We show that the mean square error decays at a rate $O(1/k^{1/4-\epsilon})$, where $\epsilon>0$ is arbitrarily small. We also show almost sure convergence of iterates to the set of fixed points. We show the applicability of our framework by applying our results to minimax optimization, linear stochastic approximation, and Lagrangian optimization., Comment: Submitted to SIAM Journal on Control and Optimization

Early management and better clinical outcomes for epileptic patients depend on seizure prediction. The accuracy and false alarm rates of existing systems are often compromised by their dependence on static thresholds and basic Electroencephalogram (EEG) properties. A novel Recurrent Neural Network (RNN)-based method for seizure start prediction is proposed in the article to overcome these limitations. As opposed to conventional techniques, the proposed system makes use of Long Short-Term Memory (LSTM) networks to extract temporal correlations from unprocessed EEG data. It enables the system to adapt dynamically to the unique EEG patterns of each patient, improving prediction accuracy. The methodology of the system comprises thorough data collecting, preprocessing, and LSTM-based feature extraction. Annotated EEG datasets are then used for model training and validation. Results show a considerable reduction in false alarm rates (average of 6.8%) and an improvement in prediction accuracy (90.2% sensitivity, 88.9% specificity, and AUC-ROC of 93). Additionally, computational efficiency is significantly higher than that of existing systems (12 ms processing time, 45 MB memory consumption). About improving seizure prediction reliability, these results demonstrate the effectiveness of the proposed RNN-based strategy, opening up possibilities for its practical application to improve epilepsy treatment.

This paper presents the multiplicity of neutral and charged hadrons produced in 90 GeV$/c$ proton-carbon interactions from a dataset taken by the NA61/SHINE experiment in 2017. Particle identification via dE/dx was performed for the charged hadrons $\pi^\pm$, $K^\pm$, and $p / \bar{p}$; the neutral hadrons $K^0_S$, $\Lambda$, and $\bar{\Lambda}$ were identified via an invariant mass analysis of their decays to charged hadrons. Double-differential multiplicity results as a function of laboratory momentum and polar angle are presented for each particle species; these results provide vital constraints on the predicted neutrino beam flux for current and future long-baseline neutrino oscillation experiments.

Data heterogeneity has been a long-standing bottleneck in studying the convergence rates of Federated Learning algorithms. In order to better understand the issue of data heterogeneity, we study the convergence rate of the Expectation-Maximization (EM) algorithm for the Federated Mixture of $K$ Linear Regressions model. We fully characterize the convergence rate of the EM algorithm under all regimes of $m/n$ where $m$ is the number of clients and $n$ is the number of data points per client. We show that with a signal-to-noise-ratio (SNR) of order $\Omega(\sqrt{K})$, the well-initialized EM algorithm converges within the minimax distance of the ground truth under each of the regimes. Interestingly, we identify that when $m$ grows exponentially in $n$, the EM algorithm only requires a constant number of iterations to converge. We perform experiments on synthetic datasets to illustrate our results. Surprisingly, the results show that rather than being a bottleneck, data heterogeneity can accelerate the convergence of federated learning algorithms.

From incorporating LLMs in education, to identifying new drugs and improving ways to charge batteries, innovators constantly try new strategies in search of better long-term outcomes for students, patients and consumers. One major bottleneck in this innovation cycle is the amount of time it takes to observe the downstream effects of a decision policy that incorporates new interventions. The key question is whether we can quickly evaluate long-term outcomes of a new decision policy without making long-term observations. Organizations often have access to prior data about past decision policies and their outcomes, evaluated over the full horizon of interest. Motivated by this, we introduce a new setting for short-long policy evaluation for sequential decision making tasks. Our proposed methods significantly outperform prior results on simulators of HIV treatment, kidney dialysis and battery charging. We also demonstrate that our methods can be useful for applications in AI safety by quickly identifying when a new decision policy is likely to have substantially lower performance than past policies., Comment: Added references for related work

Consider a game where the players' utility functions include a reward function and a linear term for each dimension, with coefficients that are controlled by the manager. We assume that the game is strongly monotone, so gradient play converges to a unique Nash equilibrium (NE). The NE is typically globally inefficient. The global performance at NE can be improved by imposing linear constraints on the NE. We therefore want the manager to pick the controlled coefficients that impose the desired constraint on the NE. However, this requires knowing the players' reward functions and action sets. Obtaining this game information is infeasible in a large-scale network and violates user privacy. To overcome this, we propose a simple algorithm that learns to shift the NE to meet the linear constraints by adjusting the controlled coefficients online. Our algorithm only requires the linear constraints violation as feedback and does not need to know the reward functions or the action sets. We prove that our algorithm converges with probability 1 to the set of NE that satisfy target linear constraints. We then prove an L2 convergence rate of near-$O(t^{-1/4})$., Comment: Submitted to IEEE Transactions on Control of Network Systems

To better align Large Language Models (LLMs) with human judgment, Reinforcement Learning from Human Feedback (RLHF) learns a reward model and then optimizes it using regularized RL. Recently, direct alignment methods were introduced to learn such a fine-tuned model directly from a preference dataset without computing a proxy reward function. These methods are built upon contrastive losses involving the log-likelihood of (dis)preferred completions according to the trained model. However, completions have various lengths, and the log-likelihood is not length-invariant. On the other side, the cross-entropy loss used in supervised training is length-invariant, as batches are typically averaged token-wise. To reconcile these approaches, we introduce a principled approach for making direct alignment length-invariant. Formally, we introduce a new averaging operator, to be composed with the optimality operator giving the best policy for the underlying RL problem. It translates into averaging the log-likelihood within the loss. We empirically study the effect of such averaging, observing a trade-off between the length of generations and their scores.

Reinforcement Learning (RL) has been used to finetune Large Language Models (LLMs) using a reward model trained from preference data, to better align with human judgment. The recently introduced direct alignment methods, which are often simpler, more stable, and computationally lighter, can more directly achieve this. However, these approaches cannot optimize arbitrary rewards, and the preference-based ones are not the only rewards of interest for LLMs (eg., unit tests for code generation or textual entailment for summarization, among others). RL-finetuning is usually done with a variation of policy gradient, which calls for on-policy or near-on-policy samples, requiring costly generations. We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data. It can be seen as an off-policy policy gradient approach that does not rely on important sampling techniques and highlights the importance of using (the right) state baseline. We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient. We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task, using a learned reward function considered as ground truth for the purpose of the experiments., Comment: EMNLP 2024

We develop a remote patient monitoring (RPM) service architecture, which has two tiers of monitoring: ordinary and intensive. The patient's health state improves or worsens in each time period according to certain probabilities, which depend on the monitoring tier. The patient incurs a "loss of quality of life" cost or an "invasiveness" cost, which is higher under intensive monitoring than under ordinary. On the other hand, their health improves faster under intensive monitoring than under ordinary. In each period, the service decides which monitoring tier to use, based on the health of the patient. We investigate the optimal policy for making that choice by formulating the problem using dynamic programming. We first provide analytic conditions for selecting ordinary vs intensive monitoring in the asymptotic regime where the number of health states is large. In the general case, we investigate the optimal policy numerically. We observe a threshold behavior, that is, when the patient's health drops below a certain threshold the service switches them to intensive monitoring, while ordinary monitoring is used during adequately good health states of the patient. The modeling and analysis provides a general framework for managing RPM services for various health conditions with medically/clinically defined system parameters., Comment: Accepted for presentation at IEEE Healthcom 2024. 7 pages

Retrieval-Augmented Generation (RAG) has revolutionized natural language processing by dynamically integrating external knowledge into Large Language Models (LLMs), addressing their limitation of static training datasets. Recent implementations of RAG leverage hierarchical data structures, which organize documents at various levels of summarization and information density. This complexity, however, can cause LLMs to "choke" on information overload, necessitating more sophisticated querying mechanisms. In this context, we introduce Hierarchical Information Retrieval Optimization (HIRO), a novel querying approach that employs a Depth-First Search (DFS)-based recursive similarity score calculation and branch pruning. This method uniquely minimizes the context delivered to the LLM without informational loss, effectively managing the challenge of excessive data. HIRO's refined approach is validated by a 10.85% improvement in performance on the NarrativeQA dataset.

Offline policy evaluation (OPE) allows us to evaluate and estimate a new sequential decision-making policy's performance by leveraging historical interaction data collected from other policies. Evaluating a new policy online without a confident estimate of its performance can lead to costly, unsafe, or hazardous outcomes, especially in education and healthcare. Several OPE estimators have been proposed in the last decade, many of which have hyperparameters and require training. Unfortunately, choosing the best OPE algorithm for each task and domain is still unclear. In this paper, we propose a new algorithm that adaptively blends a set of OPE estimators given a dataset without relying on an explicit selection using a statistical procedure. We prove that our estimator is consistent and satisfies several desirable properties for policy evaluation. Additionally, we demonstrate that when compared to alternative approaches, our estimator can be used to select higher-performing policies in healthcare and robotics. Our work contributes to improving ease of use for a general-purpose, estimator-agnostic, off-policy evaluation framework for offline RL., Comment: 22 pages

Large language models (LLMs) are quickly being adopted in a wide range of learning experiences, especially via ubiquitous and broadly accessible chat interfaces like ChatGPT and Copilot. This type of interface is readily available to students and teachers around the world, yet relatively little research has been done to assess the impact of such generic tools on student learning. Coding education is an interesting test case, both because LLMs have strong performance on coding tasks, and because LLM-powered support tools are rapidly becoming part of the workflow of professional software engineers. To help understand the impact of generic LLM use on coding education, we conducted a large-scale randomized control trial with 5,831 students from 146 countries in an online coding class in which we provided some students with access to a chat interface with GPT-4. We estimate positive benefits on exam performance for adopters, the students who used the tool, but over all students, the advertisement of GPT-4 led to a significant average decrease in exam participation. We observe similar decreases in other forms of course engagement. However, this decrease is modulated by the student's country of origin. Offering access to LLMs to students from low human development index countries increased their exam participation rate on average. Our results suggest there may be promising benefits to using LLMs in an introductory coding class, but also potential harms for engagement, which makes their longer term impact on student success unclear. Our work highlights the need for additional investigations to help understand the potential impact of future adoption and integration of LLMs into classrooms., Comment: 32 pages

A/B tests are often required to be conducted on subjects that might have social connections. For e.g., experiments on social media, or medical and social interventions to control the spread of an epidemic. In such settings, the SUTVA assumption for randomized-controlled trials is violated due to network interference, or spill-over effects, as treatments to group A can potentially also affect the control group B. When the underlying social network is known exactly, prior works have demonstrated how to conduct A/B tests adequately to estimate the global average treatment effect (GATE). However, in practice, it is often impossible to obtain knowledge about the exact underlying network. In this paper, we present UNITE: a novel estimator that relax this assumption and can identify GATE while only relying on knowledge of the superset of neighbors for any subject in the graph. Through theoretical analysis and extensive experiments, we show that the proposed approach performs better in comparison to standard estimators., Comment: AISTATS 2024

A long line of works characterizes the sample complexity of regret minimization in sequential decision-making by min-max programs. In the corresponding saddle-point game, the min-player optimizes the sampling distribution against an adversarial max-player that chooses confusing models leading to large regret. The most recent instantiation of this idea is the decision-estimation coefficient (DEC), which was shown to provide nearly tight lower and upper bounds on the worst-case expected regret in structured bandits and reinforcement learning. By re-parametrizing the offset DEC with the confidence radius and solving the corresponding min-max program, we derive an anytime variant of the Estimation-To-Decisions (E2D) algorithm. Importantly, the algorithm optimizes the exploration-exploitation trade-off online instead of via the analysis. Our formulation leads to a practical algorithm for finite model classes and linear feedback models. We further point out connections to the information ratio, decoupling coefficient and PAC-DEC, and numerically evaluate the performance of E2D on simple examples.

We introduce TartanAviation, an open-source multi-modal dataset focused on terminal-area airspace operations. TartanAviation provides a holistic view of the airport environment by concurrently collecting image, speech, and ADS-B trajectory data using setups installed inside airport boundaries. The datasets were collected at both towered and non-towered airfields across multiple months to capture diversity in aircraft operations, seasons, aircraft types, and weather conditions. In total, TartanAviation provides 3.1M images, 3374 hours of Air Traffic Control speech data, and 661 days of ADS-B trajectory data. The data was filtered, processed, and validated to create a curated dataset. In addition to the dataset, we also open-source the code-base used to collect and pre-process the dataset, further enhancing accessibility and usability. We believe this dataset has many potential use cases and would be particularly vital in allowing AI and machine learning technologies to be integrated into air traffic control systems and advance the adoption of autonomous aircraft in the airspace., Comment: 8 pages, 6 figures, 2 tables

We derive a concentration bound of the type `for all $n \geq n_0$ for some $n_0$' for TD(0) with linear function approximation. We work with online TD learning with samples from a single sample path of the underlying Markov chain. This makes our analysis significantly different from offline TD learning or TD learning with access to independent samples from the stationary distribution of the Markov chain. We treat TD(0) as a contractive stochastic approximation algorithm, with both martingale and Markov noises. Markov noise is handled using the Poisson equation and the lack of almost sure guarantees on boundedness of iterates is handled using the concept of relaxed concentration inequalities., Comment: Submitted to Stochastic Systems

Indirect experiments provide a valuable framework for estimating treatment effects in situations where conducting randomized control trials (RCTs) is impractical or unethical. Unlike RCTs, indirect experiments estimate treatment effects by leveraging (conditional) instrumental variables, enabling estimation through encouragement and recommendation rather than strict treatment assignment. However, the sample efficiency of such estimators depends not only on the inherent variability in outcomes but also on the varying compliance levels of users with the instrumental variables and the choice of estimator being used, especially when dealing with numerous instrumental variables. While adaptive experiment design has a rich literature for direct experiments, in this paper we take the initial steps towards enhancing sample efficiency for indirect experiments by adaptively designing a data collection policy over instrumental variables. Our main contribution is a practical computational procedure that utilizes influence functions to search for an optimal data collection policy, minimizing the mean-squared error of the desired (non-linear) estimator. Through experiments conducted in various domains inspired by real-world applications, we showcase how our method can significantly improve the sample efficiency of indirect experiments.

Designing reward functions for efficiently guiding reinforcement learning (RL) agents toward specific behaviors is a complex task. This is challenging since it requires the identification of reward structures that are not sparse and that avoid inadvertently inducing undesirable behaviors. Naively modifying the reward structure to offer denser and more frequent feedback can lead to unintended outcomes and promote behaviors that are not aligned with the designer's intended goal. Although potential-based reward shaping is often suggested as a remedy, we systematically investigate settings where deploying it often significantly impairs performance. To address these issues, we introduce a new framework that uses a bi-level objective to learn \emph{behavior alignment reward functions}. These functions integrate auxiliary rewards reflecting a designer's heuristics and domain knowledge with the environment's primary rewards. Our approach automatically determines the most effective way to blend these types of feedback, thereby enhancing robustness against heuristic reward misspecification. Remarkably, it can also adapt an agent's policy optimization process to mitigate suboptimalities resulting from limitations and biases inherent in the underlying RL algorithms. We evaluate our method's efficacy on a diverse set of tasks, from small-scale experiments to high-dimensional control challenges. We investigate heuristic auxiliary rewards of varying quality -- some of which are beneficial and others detrimental to the learning process. Our results show that our framework offers a robust and principled way to integrate designer-specified heuristics. It not only addresses key shortcomings of existing approaches but also consistently leads to high-performing solutions, even when given misaligned or poorly-specified auxiliary reward functions., Comment: (Spotlight) Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023)

Self-supervised learning (SSL) for clinical time series data has received significant attention in recent literature, since these data are highly rich and provide important information about a patient's physiological state. However, most existing SSL methods for clinical time series are limited in that they are designed for unimodal time series, such as a sequence of structured features (e.g., lab values and vitals signs) or an individual high-dimensional physiological signal (e.g., an electrocardiogram). These existing methods cannot be readily extended to model time series that exhibit multimodality, with structured features and high-dimensional data being recorded at each timestep in the sequence. In this work, we address this gap and propose a new SSL method -- Sequential Multi-Dimensional SSL -- where a SSL loss is applied both at the level of the entire sequence and at the level of the individual high-dimensional data points in the sequence in order to better capture information at both scales. Our strategy is agnostic to the specific form of loss function used at each level -- it can be contrastive, as in SimCLR, or non-contrastive, as in VICReg. We evaluate our method on two real-world clinical datasets, where the time series contains sequences of (1) high-frequency electrocardiograms and (2) structured data from lab values and vitals signs. Our experimental results indicate that pre-training with our method and then fine-tuning on downstream tasks improves performance over baselines on both datasets, and in several settings, can lead to improvements across different self-supervised loss functions., Comment: ICML 2023

Large transformer models trained on diverse datasets have shown a remarkable ability to learn in-context, achieving high few-shot performance on tasks they were not explicitly trained to solve. In this paper, we study the in-context learning capabilities of transformers in decision-making problems, i.e., reinforcement learning (RL) for bandits and Markov decision processes. To do so, we introduce and study Decision-Pretrained Transformer (DPT), a supervised pretraining method where the transformer predicts an optimal action given a query state and an in-context dataset of interactions, across a diverse set of tasks. This procedure, while simple, produces a model with several surprising capabilities. We find that the pretrained transformer can be used to solve a range of RL problems in-context, exhibiting both exploration online and conservatism offline, despite not being explicitly trained to do so. The model also generalizes beyond the pretraining distribution to new tasks and automatically adapts its decision-making strategies to unknown structure. Theoretically, we show DPT can be viewed as an efficient implementation of Bayesian posterior sampling, a provably sample-efficient RL algorithm. We further leverage this connection to provide guarantees on the regret of the in-context algorithm yielded by DPT, and prove that it can learn faster than algorithms used to generate the pretraining data. These results suggest a promising yet simple path towards instilling strong in-context decision-making abilities in transformers.

We study the problem of fair sequential decision making given voter preferences. In each round, a decision rule must choose a decision from a set of alternatives where each voter reports which of these alternatives they approve. Instead of going with the most popular choice in each round, we aim for proportional representation. We formalize this aim using axioms based on Proportional Justified Representation (PJR), which were proposed in the literature on multi-winner voting and were recently adapted to multi-issue decision making. The axioms require that every group of $\alpha\%$ of the voters, if it agrees in every round (i.e., approves a common alternative), then those voters must approve at least $\alpha\%$ of the decisions. A stronger version of the axioms requires that every group of $\alpha\%$ of the voters that agrees in a $\beta$ fraction of rounds must approve $\beta\cdot\alpha\%$ of the decisions. We show that three attractive voting rules satisfy axioms of this style. One of them (Sequential Phragm\'en) makes its decisions online, and the other two satisfy strengthened versions of the axioms but make decisions semi-online (Method of Equal Shares) or fully offline (Proportional Approval Voting). The first two are polynomial-time computable, and the latter is based on an NP-hard optimization, but it admits a polynomial-time local search algorithm that satisfies the same axiomatic properties. We present empirical results about the performance of these rules based on synthetic data and U.S. political elections. We also run experiments where votes are cast by preference models trained on user responses from the moral machine dataset about ethical dilemmas., Comment: 35 pages