Searcharxiv⌕ Search

arXiv subjects

Nathan Kallus

Publications and source records attributed to Nathan Kallus.

At least 91 records · Page 5Linked to original sources
arXiv2024

Near-Optimal Non-Parametric Sequential Tests and Confidence Sequences with Possibly Dependent Observations

Sequential tests and their implied confidence sequences, which are valid at arbitrary stopping times, promise flexible statistical inference and on-the-fly decision making. However, strong guarantees are limited to parametric sequential tests that under-cover in practice or concentration-bound-based sequences that over-cover and have suboptimal rejection times. In this work, we consider classic delayed-start normal-mixture sequential probability ratio tests, and we provide the first asymptotic type-I-error and expected-rejection-time guarantees under general non-parametric data generating processes, where the asymptotics are indexed by the test's burn-in time. The type-I-error results primarily leverage a martingale strong invariance principle and establish that these tests (and their implied confidence sequences) have type-I error rates asymptotically equivalent to the desired (possibly varying) $α$-level. The expected-rejection-time results primarily leverage an identity inspired by Itô's lemma and imply that, in certain asymptotic regimes, the expected rejection time is asymptotically equivalent to the minimum possible among $α$-level tests. We show how to apply our results to sequential inference on parameters defined by estimating equations, such as average treatment effects. Together, our results establish these (ostensibly parametric) tests as general-purpose, non-parametric, and near-optimal. We illustrate this via numerical simulations and a real-data application to A/B testing at Netflix.

stat.ME↗
arXiv2024

Is Cosine-Similarity of Embeddings Really About Similarity?

Cosine-similarity is the cosine of the angle between two vectors, or equivalently the dot product between their normalizations. A popular application is to quantify semantic similarity between high-dimensional objects by applying cosine-similarity to a learned low-dimensional feature embedding. This can work better but sometimes also worse than the unnormalized dot-product between embedded vectors in practice. To gain insight into this empirical observation, we study embeddings derived from regularized linear models, where closed-form solutions facilitate analytical insights. We derive analytically how cosine-similarity can yield arbitrary and therefore meaningless `similarities.' For some linear models the similarities are not even unique, while for others they are implicitly controlled by the regularization. We discuss implications beyond linear models: a combination of different regularizations are employed when learning deep models; these have implicit and unintended effects when taking cosine-similarities of the resulting embeddings, rendering results opaque and possibly arbitrary. Based on these insights, we caution against blindly using cosine-similarity and outline alternatives.

cs.IR↗
arXiv2024

Applied Causal Inference Powered by ML and AI

An introduction to the emerging fusion of machine learning and causal inference. The book presents ideas from classical structural equation models (SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and structural causal models (SCMs), and covers Double/Debiased Machine Learning methods to do inference in such models using modern predictive tools.

econ.EM↗
arXiv2024

More Benefits of Being Distributional: Second-Order Bounds for Reinforcement Learning

In this paper, we prove that Distributional Reinforcement Learning (DistRL), which learns the return distribution, can obtain second-order bounds in both online and offline RL in general settings with function approximation. Second-order bounds are instance-dependent bounds that scale with the variance of return, which we prove are tighter than the previously known small-loss bounds of distributional RL. To the best of our knowledge, our results are the first second-order bounds for low-rank MDPs and for offline RL. When specializing to contextual bandits (one-step RL problem), we show that a distributional learning based optimism algorithm achieves a second-order worst-case regret bound, and a second-order gap dependent bound, simultaneously. We also empirically demonstrate the benefit of DistRL in contextual bandits on real-world datasets. We highlight that our analysis with DistRL is relatively simple, follows the general framework of optimism in the face of uncertainty and does not require weighted regression. Our results suggest that DistRL is a promising framework for obtaining second-order bounds in general RL settings, thus further reinforcing the benefits of DistRL.

cs.LG↗
arXiv2024

Evaluating the Surrogate Index as a Decision-Making Tool Using 200 A/B Tests at Netflix

Surrogate index approaches have recently become a popular method of estimating longer-term impact from shorter-term outcomes. In this paper, we leverage 1098 test arms from 200 A/B tests at Netflix to empirically investigate to what degree would decisions made using a surrogate index utilizing 14 days of data would align with those made using direct measurement of day 63 treatment effects. Focusing specifically on linear "auto-surrogate" models that utilize the shorter-term observations of the long-term outcome of interest, we find that the statistical inferences that we would draw from using the surrogate index are ~95% consistent with those from directly measuring the long-term treatment effect. Moreover, when we restrict ourselves to the set of tests that would be "launched" (i.e. positive and statistically significant) based on the 63-day directly measured treatment effects, we find that relying instead on the surrogate index achieves 79% and 65% recall.

stat.AP↗
arXiv2023

Future-Dependent Value-Based Off-Policy Evaluation in POMDPs

We study off-policy evaluation (OPE) for partially observable MDPs (POMDPs) with general function approximation. Existing methods such as sequential importance sampling estimators and fitted-Q evaluation suffer from the curse of horizon in POMDPs. To circumvent this problem, we develop a novel model-free OPE method by introducing future-dependent value functions that take future proxies as inputs. Future-dependent value functions play similar roles as classical value functions in fully-observable MDPs. We derive a new Bellman equation for future-dependent value functions as conditional moment equations that use history proxies as instrumental variables. We further propose a minimax learning method to learn future-dependent value functions using the new Bellman equation. We obtain the PAC result, which implies our OPE estimator is consistent as long as futures and histories contain sufficient information about latent states, and the Bellman completeness. Finally, we extend our methods to learning of dynamics and establish the connection between our approach and the well-known spectral learning methods in POMDPs.

cs.LG↗
arXiv2023

Offline Minimax Soft-Q-learning Under Realizability and Partial Coverage

In offline reinforcement learning (RL) we have no opportunity to explore so we must make assumptions that the data is sufficient to guide picking a good policy, taking the form of assuming some coverage, realizability, Bellman completeness, and/or hard margin (gap). In this work we propose value-based algorithms for offline RL with PAC guarantees under just partial coverage, specifically, coverage of just a single comparator policy, and realizability of soft (entropy-regularized) Q-function of the single policy and a related function defined as a saddle point of certain minimax optimization problem. This offers refined and generally more lax conditions for offline RL. We further show an analogous result for vanilla Q-functions under a soft margin condition. To attain these guarantees, we leverage novel minimax learning algorithms to accurately estimate soft or vanilla Q-functions with $L^2$-convergence guarantees. Our algorithms' loss functions arise from casting the estimation problems as nonlinear convex optimization problems and Lagrangifying.

cs.LG↗
arXiv2023

Long-Term Causal Inference with Imperfect Surrogates using Many Weak Experiments, Proxies, and Cross-Fold Moments

Inferring causal effects on long-term outcomes using short-term surrogates is crucial to rapid innovation. However, even when treatments are randomized and surrogates fully mediate their effect on outcomes, it's possible that we get the direction of causal effects wrong due to confounding between surrogates and outcomes -- a situation famously known as the surrogate paradox. The availability of many historical experiments offer the opportunity to instrument for the surrogate and bypass this confounding. However, even as the number of experiments grows, two-stage least squares has non-vanishing bias if each experiment has a bounded size, and this bias is exacerbated when most experiments barely move metrics, as occurs in practice. We show how to eliminate this bias using cross-fold procedures, JIVE being one example, and construct valid confidence intervals for the long-term effect in new experiments where long-term outcome has not yet been observed. Our methodology further allows to proxy for effects not perfectly mediated by the surrogates, allowing us to handle both confounding and effect leakage as violations of standard statistical surrogacy conditions.

stat.ME↗
arXiv2023

Off-Policy Evaluation for Large Action Spaces via Policy Convolution

Developing accurate off-policy estimators is crucial for both evaluating and optimizing for new policies. The main challenge in off-policy estimation is the distribution shift between the logging policy that generates data and the target policy that we aim to evaluate. Typically, techniques for correcting distribution shift involve some form of importance sampling. This approach results in unbiased value estimation but often comes with the trade-off of high variance, even in the simpler case of one-step contextual bandits. Furthermore, importance sampling relies on the common support assumption, which becomes impractical when the action space is large. To address these challenges, we introduce the Policy Convolution (PC) family of estimators. These methods leverage latent structure within actions -- made available through action embeddings -- to strategically convolve the logging and target policies. This convolution introduces a unique bias-variance trade-off, which can be controlled by adjusting the amount of convolution. Our experiments on synthetic and benchmark datasets demonstrate remarkable mean squared error (MSE) improvements when using PC, especially when either the action space or policy mismatch becomes large, with gains of up to 5 - 6 orders of magnitude over existing estimators.

cs.LG↗
arXiv2023

JoinGym: An Efficient Query Optimization Environment for Reinforcement Learning

Join order selection (JOS) is the problem of ordering join operations to minimize total query execution cost and it is the core NP-hard combinatorial optimization problem of query optimization. In this paper, we present JoinGym, a lightweight and easy-to-use query optimization environment for reinforcement learning (RL) that captures both the left-deep and bushy variants of the JOS problem. Compared to existing query optimization environments, the key advantages of JoinGym are usability and significantly higher throughput which we accomplish by simulating query executions entirely offline. Under the hood, JoinGym simulates a query plan's cost by looking up intermediate result cardinalities from a pre-computed dataset. We release a novel cardinality dataset for $3300$ SQL queries based on real IMDb workloads which may be of independent interest, e.g., for cardinality estimation. Finally, we extensively benchmark four RL algorithms and find that their cost distributions are heavy-tailed, which motivates future work in risk-sensitive RL. In sum, JoinGym enables users to rapidly prototype RL algorithms on realistic database problems without needing to setup and run live systems.

cs.LG↗
arXiv2023

Provable Offline Preference-Based Reinforcement Learning

In this paper, we investigate the problem of offline Preference-based Reinforcement Learning (PbRL) with human feedback where feedback is available in the form of preference between trajectory pairs rather than explicit rewards. Our proposed algorithm consists of two main steps: (1) estimate the implicit reward using Maximum Likelihood Estimation (MLE) with general function approximation from offline data and (2) solve a distributionally robust planning problem over a confidence set around the MLE. We consider the general reward setting where the reward can be defined over the whole trajectory and provide a novel guarantee that allows us to learn any target policy with a polynomial number of samples, as long as the target policy is covered by the offline data. This guarantee is the first of its kind with general function approximation. To measure the coverage of the target policy, we introduce a new single-policy concentrability coefficient, which can be upper bounded by the per-trajectory concentrability coefficient. We also establish lower bounds that highlight the necessity of such concentrability and the difference from standard RL, where state-action-wise rewards are directly observed. We further extend and analyze our algorithm when the feedback is given over action pairs.

cs.LG↗
arXiv2023

The Benefits of Being Distributional: Small-Loss Bounds for Reinforcement Learning

While distributional reinforcement learning (DistRL) has been empirically effective, the question of when and why it is better than vanilla, non-distributional RL has remained unanswered. This paper explains the benefits of DistRL through the lens of small-loss bounds, which are instance-dependent bounds that scale with optimal achievable cost. Particularly, our bounds converge much faster than those from non-distributional approaches if the optimal cost is small. As warmup, we propose a distributional contextual bandit (DistCB) algorithm, which we show enjoys small-loss regret bounds and empirically outperforms the state-of-the-art on three real-world tasks. In online RL, we propose a DistRL algorithm that constructs confidence sets using maximum likelihood estimation. We prove that our algorithm enjoys novel small-loss PAC bounds in low-rank MDPs. As part of our analysis, we introduce the $\ell_1$ distributional eluder dimension which may be of independent interest. Then, in offline RL, we show that pessimistic DistRL enjoys small-loss PAC bounds that are novel to the offline setting and are more robust to bad single-policy coverage.

cs.LG↗
arXiv2023

Large Language Models as Zero-Shot Conversational Recommenders

In this paper, we present empirical studies on conversational recommendation tasks using representative large language models in a zero-shot setting with three primary contributions. (1) Data: To gain insights into model behavior in "in-the-wild" conversational recommendation scenarios, we construct a new dataset of recommendation-related conversations by scraping a popular discussion website. This is the largest public real-world conversational recommendation dataset to date. (2) Evaluation: On the new dataset and two existing conversational recommendation datasets, we observe that even without fine-tuning, large language models can outperform existing fine-tuned conversational recommendation models. (3) Analysis: We propose various probing tasks to investigate the mechanisms behind the remarkable performance of large language models in conversational recommendation. We analyze both the large language models' behaviors and the characteristics of the datasets, providing a holistic understanding of the models' effectiveness, limitations and suggesting directions for the design of future conversational recommenders

cs.IR↗
arXiv2023

Generalization Bounds and Representation Learning for Estimation of Potential Outcomes and Causal Effects

Practitioners in diverse fields such as healthcare, economics and education are eager to apply machine learning to improve decision making. The cost and impracticality of performing experiments and a recent monumental increase in electronic record keeping has brought attention to the problem of evaluating decisions based on non-experimental observational data. This is the setting of this work. In particular, we study estimation of individual-level causal effects, such as a single patient's response to alternative medication, from recorded contexts, decisions and outcomes. We give generalization bounds on the error in estimated effects based on distance measures between groups receiving different treatments, allowing for sample re-weighting. We provide conditions under which our bound is tight and show how it relates to results for unsupervised domain adaptation. Led by our theoretical results, we devise representation learning algorithms that minimize our bound, by regularizing the representation's induced treatment group distance, and encourage sharing of information between treatment groups. We extend these algorithms to simultaneously learn a weighted representation to further reduce treatment group distances. Finally, an experimental evaluation on real and synthetic data shows the value of our proposed representation architecture and regularization scheme.

cs.LG↗
arXiv2023

Source Condition Double Robust Inference on Functionals of Inverse Problems

We consider estimation of parameters defined as linear functionals of solutions to linear inverse problems. Any such parameter admits a doubly robust representation that depends on the solution to a dual linear inverse problem, where the dual solution can be thought as a generalization of the inverse propensity function. We provide the first source condition double robust inference method that ensures asymptotic normality around the parameter of interest as long as either the primal or the dual inverse problem is sufficiently well-posed, without knowledge of which inverse problem is the more well-posed one. Our result is enabled by novel guarantees for iterated Tikhonov regularized adversarial estimators for linear inverse problems, over general hypothesis spaces, which are developments of independent interest.

stat.ME↗
arXiv2023

Fast Rates for the Regret of Offline Reinforcement Learning

We study the regret of reinforcement learning from offline data generated by a fixed behavior policy in an infinite-horizon discounted Markov decision process (MDP). While existing analyses of common approaches, such as fitted $Q$-iteration (FQI), suggest a $O(1/\sqrt{n})$ convergence for regret, empirical behavior exhibits \emph{much} faster convergence. In this paper, we present a finer regret analysis that exactly characterizes this phenomenon by providing fast rates for the regret convergence. First, we show that given any estimate for the optimal quality function $Q^*$, the regret of the policy it defines converges at a rate given by the exponentiation of the $Q^*$-estimate's pointwise convergence rate, thus speeding it up. The level of exponentiation depends on the level of noise in the \emph{decision-making} problem, rather than the estimation problem. We establish such noise levels for linear and tabular MDPs as examples. Second, we provide new analyses of FQI and Bellman residual minimization to establish the correct pointwise convergence guarantees. As specific cases, our results imply $O(1/n)$ regret rates in linear cases and $\exp(-Ω(n))$ regret rates in tabular cases. We extend our findings to general function approximation by extending our results to regret guarantees based on $L_p$-convergence rates for estimating $Q^*$ rather than pointwise rates, where $L_2$ guarantees for nonparametric $Q^*$-estimation can be ensured under mild conditions.

cs.LG↗
arXiv2023

Inference on Strongly Identified Functionals of Weakly Identified Functions

In a variety of applications, including nonparametric instrumental variable (NPIV) analysis, proximal causal inference under unmeasured confounding, and missing-not-at-random data with shadow variables, we are interested in inference on a continuous linear functional (e.g., average causal effects) of nuisance function (e.g., NPIV regression) defined by conditional moment restrictions. These nuisance functions are generally weakly identified, in that the conditional moment restrictions can be severely ill-posed as well as admit multiple solutions. This is sometimes resolved by imposing strong conditions that imply the function can be estimated at rates that make inference on the functional possible. In this paper, we study a novel condition for the functional to be strongly identified even when the nuisance function is not; that is, the functional is amenable to asymptotically-normal estimation at $\sqrt{n}$-rates. The condition implies the existence of debiasing nuisance functions, and we propose penalized minimax estimators for both the primary and debiasing nuisance functions. The proposed nuisance estimators can accommodate flexible function classes, and importantly they can converge to fixed limits determined by the penalization regardless of the identifiability of the nuisances. We use the penalized nuisance estimators to form a debiased estimator for the functional of interest and prove its asymptotic normality under generic high-level conditions, which provide for asymptotically valid confidence intervals. We also illustrate our method in a novel partially linear proximal causal inference problem and a partially linear instrumental variable regression problem.

stat.ME↗
arXiv2023

An Empirical Evaluation of the Impact of New York's Bail Reform on Crime Using Synthetic Controls

We conduct an empirical evaluation of the impact of New York's bail reform on crime. New York State's Bail Elimination Act went into effect on January 1, 2020, eliminating money bail and pretrial detention for nearly all misdemeanor and nonviolent felony defendants. Our analysis of effects on aggregate crime rates after the reform informs the understanding of bail reform and general deterrence. We conduct a synthetic control analysis for a comparative case study of impact of bail reform. We focus on synthetic control analysis of post-intervention changes in crime for assault, theft, burglary, robbery, and drug crimes, constructing a dataset from publicly reported crime data of 27 large municipalities. Our findings, including placebo checks and other robustness checks, show that for assault, theft, and drug crimes, there is no significant impact of bail reform on crime; for burglary and robbery, we similarly have null findings but the synthetic control is also more variable so these are deemed less conclusive.

stat.AP↗