arXiv subjects

Peter Zhang

Publications and source records attributed to Peter Zhang.

At least 19 recordsLinked to original sources

arXiv2025

Covariate-adjusted Group Sequential Comparisons of Restricted Mean Survival Times

The restricted mean survival time (RMST) is the mean survival time in the study population followed up to a specific time point, and is simply the area under the survival curve up to the specific time point. The difference between two RMSTs quantifies the group difference in a time scale by measuring the integrated difference between survival curves in the two treatment groups. This paper develops a group sequential (GS) test of comparing two RMSTs up to a restriction time point under a stratified proportional hazards model with stratum representing treatment status. This covariate-adjusted GS test does not require the assumption of a constant hazard ratio over time between the two treatment groups and is valid whether or not the proportional hazards assumption holds for the treatment effect. We establish the large-sample properties of the covariate-adjusted GS test and show that the proposed test statistics sequentially computed at different interim analysis times possess an independent increments covariance structure. Using currently available methodology, this joint independent increments structure allows us to calculate sequential stopping boundaries for preserving the desired type I error probability and maintaining the required statistical power. We evaluate the small-sample performance of the proposed GS test via a simulation study and illustrate its real-world application through analysis of a clinical trial dataset from the BMT CTN 1101 study.

stat.ME↗

arXiv2025

Robust Paths: Geometry and Computation

Applying robust optimization often requires selecting an appropriate uncertainty set both in shape and size, a choice that directly affects the trade-off between average-case and worst-case performances. In practice, this calibration is usually done via trial-and-error: solving the robust optimization problem many times with different uncertainty set shapes and sizes, and examining their performance trade-off. This process is computationally expensive and ad hoc. In this work, we take a principled approach to study this issue for robust optimization problems with linear objective functions, convex feasible regions, and convex uncertainty sets. We introduce and study what we define as the robust path: a set of robust solutions obtained by varying the uncertainty set's parameters. Our central geometric insight is that a robust path can be characterized as a Bregman projection of a curve (whose geometry is defined by the uncertainty set) onto the feasible region. This leads to a surprising discovery that the robust path can be approximated via the trajectories of standard optimization algorithms, such as the proximal point method, of the deterministic counterpart problem. We give a sharp approximation error bound and show it depends on the geometry of the feasible region and the uncertainty set. We also illustrate two special cases where the approximation error is zero: the feasible region is polyhedrally monotone (e.g., a simplex feasible region under an ellipsoidal uncertainty set), or the feasible region and the uncertainty set follow a dual relationship. We demonstrate the practical impact of this approach in two settings: portfolio optimization and adversarial deep learning.

math.OC↗

arXiv2025

Domain Gating Ensemble Networks for AI-Generated Text Detection

As state-of-the-art language models continue to improve, the need for robust detection of machine-generated text becomes increasingly critical. However, current state-of-the-art machine text detectors struggle to adapt to new unseen domains and generative models. In this paper we present DoGEN (Domain Gating Ensemble Networks), a technique that allows detectors to adapt to unseen domains by ensembling a set of domain expert detector models using weights from a domain classifier. We test DoGEN on a wide variety of domains from leading benchmarks and find that it achieves state-of-the-art performance on in-domain detection while outperforming models twice its size on out-of-domain detection. We release our code and trained models to assist in future research in domain-adaptive AI detection.

cs.CL↗

arXiv2025

Visual and Textual Prompts in VLLMs for Enhancing Emotion Recognition

Vision Large Language Models (VLLMs) exhibit promising potential for multi-modal understanding, yet their application to video-based emotion recognition remains limited by insufficient spatial and contextual awareness. Traditional approaches, which prioritize isolated facial features, often neglect critical non-verbal cues such as body language, environmental context, and social interactions, leading to reduced robustness in real-world scenarios. To address this gap, we propose Set-of-Vision-Text Prompting (SoVTP), a novel framework that enhances zero-shot emotion recognition by integrating spatial annotations (e.g., bounding boxes, facial landmarks), physiological signals (facial action units), and contextual cues (body posture, scene dynamics, others' emotions) into a unified prompting strategy. SoVTP preserves holistic scene information while enabling fine-grained analysis of facial muscle movements and interpersonal dynamics. Extensive experiments show that SoVTP achieves substantial improvements over existing visual prompting methods, demonstrating its effectiveness in enhancing VLLMs' video emotion recognition capabilities.

cs.CV↗

arXiv2025

A Gauge Set Framework for Flexible Robustness Design

This paper proposes a unified framework for designing robustness in optimization under uncertainty using gauge sets, convex sets that generalize distance and capture how distributions may deviate from a nominal reference. Representing robustness through a gauge set reweighting formulation brings many classical robustness paradigms under a single convex-analytic perspective. The corresponding dual problem, the upper approximator regularization model, reveals a direct connection between distributional perturbations and objective regularization via polar gauge sets. This framework decouples the design of the nominal distribution, distance metric, and reformulation method, components often entangled in classical approaches, thus enabling modular and composable robustness modeling. We further provide a gauge set algebra toolkit that supports intersection, summation, convex combination, and composition, enabling complex ambiguity structures to be assembled from simpler components. For computational tractability under continuously supported uncertainty, we introduce two general finite-dimensional reformulation methods. The functional parameterization approach guarantees any prescribed gauge-based robustness through flexible selection of function bases, while the envelope representation approach yields exact reformulations under empirical nominal distributions and is asymptotically exact for arbitrary nominal choices. A detailed case study demonstrates how the framework accommodates diverse robustness requirements while admitting multiple tractable reformulations.

math.OC↗

arXiv2024

Optimal probabilistic feature shifts for reclassification in tree ensembles

In this paper we provide a novel mathematical optimization based methodology to perturb the features of a given observation to be re-classified, by a tree ensemble classification rule, to a certain desired class. The method is based on these facts: the most viable changes for an observation to reach the desired class do not always coincide with the closest distance point (in the feature space) of the target class; individuals put effort on a few number of features to reach the desired class; and each individual is endowed with a probability to change each of its features to a given value, which determines the overall probability of changing to the target class. Putting all together, we provide different methods to find the features where the individuals must exert effort to maximize the probability to reach the target class. Our method also allows us to rank the most important features in the tree-ensemble. The proposed methodology is tested on a real dataset, validating the proposal.

math.OC↗

arXiv2024

OPTIMA: Optimized Policy for Intelligent Multi-Agent Systems Enables Coordination-Aware Autonomous Vehicles

Coordination among connected and autonomous vehicles (CAVs) is advancing due to developments in control and communication technologies. However, much of the current work is based on oversimplified and unrealistic task-specific assumptions, which may introduce vulnerabilities. This is critical because CAVs not only interact with their environment but are also integral parts of it. Insufficient exploration can result in policies that carry latent risks, highlighting the need for methods that explore the environment both extensively and efficiently. This work introduces OPTIMA, a novel distributed reinforcement learning framework for cooperative autonomous vehicle tasks. OPTIMA alternates between thorough data sampling from environmental interactions and multi-agent reinforcement learning algorithms to optimize CAV cooperation, emphasizing both safety and efficiency. Our goal is to improve the generality and performance of CAVs in highly complex and crowded scenarios. Furthermore, the industrial-scale distributed training system easily adapts to different algorithms, reward functions, and strategies.

cs.MA↗

arXiv2024

Approximating Multiple Robust Optimization Solutions in One Pass via Proximal Point Methods

Robust optimization provides a principled and unified framework to model many problems in modern operations research and computer science applications, such as risk measures minimization and adversarially robust machine learning. To use a robust solution (e.g., to implement an investment portfolio or perform robust machine learning inference), the user has to a priori decide the trade-off between efficiency (nominal performance) and robustness (worst-case performance) of the solution by choosing the uncertainty level hyperparameters. In many applications, this amounts to solving the problem many times and comparing them, each from a different hyperparameter setting. This makes robust optimization practically cumbersome or even intractable. We present a novel procedure based on the proximal point method (PPM) to efficiently approximate many Pareto efficient robust solutions at once. This effectively reduces the total compute requirement from $N \times T$ to $2 \times T$, where $N$ is the number of robust solutions to be generated, and $T$ is the time to obtain one robust solution. We prove this procedure can produce exact Pareto efficient robust solutions for a class of robust linear optimization problems. For more general problems, we prove that with high probability, our procedure gives a good approximation of the efficiency-robustness trade-off in random robust linear optimization instances. We conduct numerical experiments to demonstrate.

math.OC↗

arXiv2024

OpenDebateEvidence: A Massive-Scale Argument Mining and Summarization Dataset

We introduce OpenDebateEvidence, a comprehensive dataset for argument mining and summarization sourced from the American Competitive Debate community. This dataset includes over 3.5 million documents with rich metadata, making it one of the most extensive collections of debate evidence. OpenDebateEvidence captures the complexity of arguments in high school and college debates, providing valuable resources for training and evaluation. Our extensive experiments demonstrate the efficacy of fine-tuning state-of-the-art large language models for argumentative abstractive summarization across various methods, models, and datasets. By providing this comprehensive resource, we aim to advance computational argumentation and support practical applications for debaters, educators, and researchers. OpenDebateEvidence is publicly available to support further research and innovation in computational argumentation. Access it here: https://huggingface.co/datasets/Hellisotherpeople/OpenDebateEvidence-Anonymized

cs.CL↗

arXiv2024

Covariate-adjusted Group Sequential Comparisons of Survival Probabilities

In confirmatory clinical trials, survival outcomes are frequently studied and interim analyses for efficacy and/or futility are often desirable. Methods such as the log rank test and Cox regression model are commonly used to compare treatments in this setting. They rely on a proportional hazards (PH) assumption and are subject to type I error rate inflation and loss of power when PH are violated. Such violations may be expected a priori, particularly when the mechanisms of treatments differ such as immunotherapy vs. chemotherapy for treating cancer. We develop group sequential tests for comparing survival curves with covariate adjustment that allow for interim analyses in the presence of non-PH and offer easily interpreted, clinically meaningful summary measures of the treatment effect. The joint distribution of repeatedly computed test statistics converges to the canonical joint distribution with a Markov structure. The asymptotic distribution of the test statistics allows marginal comparisons of survival probabilities at multiple fixed time points and facilitates both critical value specification to maintain type I error control and sample size/power determination. Simulations demonstrate that the achieved type I error rate and power of the proposed tests meet targeted levels and are robust to the PH assumption and covariate influence. The proposed tests are illustrated using a clinical trial dataset from the Blood and Marrow Transplant Clinical Trials Network 1101 trial.

stat.ME↗

arXiv2024

Binary Noise for Binary Tasks: Masked Bernoulli Diffusion for Unsupervised Anomaly Detection

The high performance of denoising diffusion models for image generation has paved the way for their application in unsupervised medical anomaly detection. As diffusion-based methods require a lot of GPU memory and have long sampling times, we present a novel and fast unsupervised anomaly detection approach based on latent Bernoulli diffusion models. We first apply an autoencoder to compress the input images into a binary latent representation. Next, a diffusion model that follows a Bernoulli noise schedule is employed to this latent space and trained to restore binary latent representations from perturbed ones. The binary nature of this diffusion model allows us to identify entries in the latent space that have a high probability of flipping their binary code during the denoising process, which indicates out-of-distribution data. We propose a masking algorithm based on these probabilities, which improves the anomaly detection scores. We achieve state-of-the-art performance compared to other diffusion-based unsupervised anomaly detection algorithms while significantly reducing sampling time and memory consumption. The code is available at https://github.com/JuliaWolleb/Anomaly_berdiff.

cs.CV↗

arXiv2023

Taking Advice from ChatGPT

A growing literature studies how humans incorporate advice from algorithms. This study examines an algorithm with millions of daily users: ChatGPT. In a preregistered study, 118 student participants answer 2,828 multiple-choice questions across 25 academic subjects. Participants receive advice from a GPT model and can update their initial responses. The advisor's identity ("AI chatbot" versus a human "expert"), presence of a written justification, and advice correctness do not significantly affect weight on advice. Instead, participants weigh advice more heavily if they (1) are unfamiliar with the topic, (2) used ChatGPT in the past, or (3) received more accurate advice previously. The last two effects -- algorithm familiarity and experience -- are stronger with an AI chatbot as the advisor. Participants that receive written justifications are able to discern correct advice and update accordingly. Student participants are miscalibrated in their judgements of ChatGPT advice accuracy; one reason is that they significantly misjudge the accuracy of ChatGPT on 11/25 topics. Participants under-weigh advice by over 50% and can score better by trusting ChatGPT more.

cs.HC↗

arXiv2023

Distributionally Robust Principal-Agent Problems and Optimality of Contracts

We propose a distributionally robust principal agent formulation, which generalizes some common variants of worst-case and Bayesian principal agent problems. We construct a theoretical framework to certify whether any surjective contract family is optimal, and bound its sub-optimality. We then apply the framework to study the optimality of affine contracts. We show with geometric intuition that these simple contract families are optimal when the surplus function is convex and there exists a technology type that is simultaneously least productive and least efficient. We also provide succinct expressions to quantify the optimality gap of any surplus function, based on its concave biconjugate. This new framework complements the current literature in two ways: invention of a new toolset; understanding affine contracts' performance in a larger landscape. Our results also shed light on the technical roots of this question: why are there more positive results in the recent literature that show simple contracts' optimality in robust settings rather than stochastic settings? This phenomenon is related to two technical facts: the sum of quasi-concave functions is not quasi-concave, and the maximization and expectation operators do not commute.

econ.TH↗

arXiv2023

Goal Driven Discovery of Distributional Differences via Language Descriptions

Mining large corpora can generate useful discoveries but is time-consuming for humans. We formulate a new task, D5, that automatically discovers differences between two large corpora in a goal-driven way. The task input is a problem comprising a research goal "$\textit{comparing the side effects of drug A and drug B}$" and a corpus pair (two large collections of patients' self-reported reactions after taking each drug). The output is a language description (discovery) of how these corpora differ (patients taking drug A "$\textit{mention feelings of paranoia}$" more often). We build a D5 system, and to quantitatively measure its performance, we 1) contribute a meta-dataset, OpenD5, aggregating 675 open-ended problems ranging across business, social sciences, humanities, machine learning, and health, and 2) propose a set of unified evaluation metrics: validity, relevance, novelty, and significance. With the dataset and the unified metrics, we confirm that language models can use the goals to propose more relevant, novel, and significant candidate discoveries. Finally, our system produces discoveries previously unknown to the authors on a wide range of applications in OpenD5, including temporal and demographic differences in discussion topics, political stances and stereotypes in speech, insights in commercial reviews, and error patterns in NLP models.

cs.CL↗

arXiv2022

Efficient Methods for Approximating the Shapley Value for Asset Sharing in Energy Communities

With the emergence of energy communities, where a number of prosumers invest in shared generation and storage, the issue of fair allocation of benefits is increasingly important. The Shapley value has attracted increasing interest for redistribution in energy settings - however, computing it exactly is intractable beyond a few dozen prosumers. In this paper, we first conduct a systematic review of the literature on the use of Shapley value in energy-related applications, as well as efforts to compute or approximate it. Next, we formalise the main methods for approximating the Shapley value in community energy settings, and propose a new one, which we call the stratified expected value approximation. To compare the performance of these methods, we design a novel method for exact Shapley value computation, which can be applied to communities of up to several hundred agents by clustering the prosumers into a smaller number of demand profiles. We perform a large-scale experimental comparison of the proposed methods, for communities of up to 200 prosumers, using large-scale, publicly available data from two large-scale energy trials in the UK (UKERC Energy Data Centre, 2017, UK Power Networks Innovation, 2021). Our analysis shows that, as the number of agents in the community increases, the relative difference to the exact Shapley value converges to under 1% for all the approximation methods considered. In particular, for most experimental scenarios, we show that there is no statistical difference between the newly proposed stratified expected value method and the existing state-of-the-art method that uses adaptive sampling (O'Brien et al., 2015), although the cost of computation for large communities is an order of magnitude lower.

cs.GT↗

arXiv2022

A Mathematical Programming Approach to Optimal Classification Forests

This paper introduces Weighted Optimal Classification Forests (WOCFs), a new family of classifiers that takes advantage of an optimal ensemble of decision trees to derive accurate and interpretable classifiers. We propose a novel mathematical optimization-based methodology which simultaneously constructs a given number of trees, each of them providing a predicted class for the observations in the feature space. The classification rule is derived by assigning to each observation its most frequently predicted class among the trees. We provide a mixed integer linear programming formulation (MIP) for the problem and several novel MIP strengthening / scaling techniques. We report the results of our computational experiments, from which we conclude that our method has equal or superior performance compared with state-of-the-art tree-based classification methods for small to medium-sized instances. We also present three real-world case studies showing that our methodology has very interesting implications in terms of interpretability. Overall, WOCFs complement existing methods such as CART, Optimal Classification Trees, Random Forests and XGBoost. In addition to its Pareto improvement on accuracy and interpretability, we also see unique properties emerging in terms of different trees focusing on different feature variables. This provides nontrivial improvement in interpretability and usability of the trained model in terms of counterfactual explanation. Thus, despite the apparent computational challenge of WOCFs that limit the size of the problems that can be efficiently solved with current MIP, this is an important research direction that can lead to qualitatively different insights for researchers and complement the toolbox of practitioners for high stakes problems.

math.OC↗

arXiv2021

Fleet Sizing and Allocation for On-demand Last-Mile Transportation Systems

The last-mile problem refers to the provision of travel service from the nearest public transportation node to home or other destination. Last-Mile Transportation Systems (LMTS), which have recently emerged, provide on-demand shared transportation. In this paper, we investigate the fleet sizing and allocation problem for the on-demand LMTS. Specifically, we consider the perspective of a last-mile service provider who wants to determine the number of servicing vehicles to allocate to multiple last-mile service regions in a particular city. In each service region, passengers demanding last-mile services arrive in batches, and allocated vehicles deliver passengers to their final destinations. The passenger demand (i.e., the size of each batch of passengers) is random and hard to predict in advance, especially with limited data during the planning process. The quality of fleet allocation decisions is a function of vehicle fixed cost plus a weighted sum of passenger's waiting time before boarding a vehicle and in-vehicle riding time. We propose and analyze two models-a stochastic programming model and a distributionally robust optimization model-to solve the problem, assuming a known and unknown distribution of the demand, respectively. We conduct extensive numerical experiments to evaluate the models and discuss insights and implications into the optimal fleet sizing and allocation for the on-demand LMTS under demand uncertainty.

math.OC↗

arXiv2021

Travel Cadence and Epidemic Spread

In this paper, we study how interactions between populations impact epidemic spread. We extend the classical SEIR model to include both integration-based disease transmission simulation and population flow. Our model differs from existing ones by having a more detailed representation of travel patterns, without losing tractability. This allows us to study the epidemic consequence of inter-regional travel with high fidelity. In particular, we define \emph{travel cadence} as a two-dimensional measure of inter-regional travel, and show that both dimensions modulate epidemic spread. This technical insight leads to policy recommendations, pointing to a family of simple policy trajectories that can effectively curb epidemic spread while maintaining a basic level of mobility.

physics.soc-ph↗