In epidemiological settings, Machine Learning (ML) is gaining popularity for hypothesis-free discovery of risk (or protective) factors. Although ML is strong at discovering non-linearities and interactions, this power is currently compromised by a lack of reliable inference. Although local measures of feature effect can be combined with tree ensembles, uncertainty quantifications for these measures remain only partially available and oftentimes unsatisfactory. We propose RuleSHAP, a framework for using rule-based, hypothesis-free discovery that combines sparse Bayesian regression, tree ensembles and Shapley values in a one-step procedure that both detects and tests complex patterns at the individual level. To ease computation, we derive a formula that computes marginal Shapley values more efficiently for our setting. We demonstrate the validity of our framework on simulated data. To illustrate, we apply our machinery to data from an epidemiological cohort to detect and infer several effects for high cholesterol and blood pressure, such as nonlinear interaction effects between features like age, sex, ethnicity, BMI and glucose level.

This article describes the process of creating a script and conducting an analytical study of a dataset using the DeepMIMO emulator. An advertorial attack was carried out using the FGSM method to maximize the gradient. A comparison is made of the effectiveness of binary classifiers in the task of detecting distorted data. The dynamics of changes in the quality indicators of the regression model were analyzed in conditions without adversarial attacks, during an adversarial attack and when the distorted data was isolated. It is shown that an adversarial FGSM attack with gradient maximization leads to an increase in the value of the MSE metric by 33% and a decrease in the R2 indicator by 10% on average. The LightGBM binary classifier effectively identifies data with adversarial anomalies with 98% accuracy. Regression machine learning models are susceptible to adversarial attacks, but rapid analysis of network traffic and data transmitted over the network makes it possible to identify malicious activity

That is, testing whether two random vectors X and Y are independent given a third random vector Z . For example, there are conditional independence tests based on conditional densities [SW08], characteristic functions [SW07], empirical likelihood ratios [SW14], discretization [Mar05; Hua10], permutation [Dor+14; Sen+17], kernels [Fuk+07; Zha+11; SP11], copulas [BRT12], and conditional mutual information [Run18b]. Also, there are many conditional independence tests based on regressing X on Z and Y on Z followed by testing for independence between the residuals [Pat+09; Pet+14; Ram14; FFX20; ZZG17; Zha+19]. Unfortunately, conditional independence tests oftentimes struggle to control the Type-I error in finite samples, as shown by Shah and Peters [SP20]. In fact, Shah and Peters [SP20] prove that conditional independence testing is fundamentally impossible without making further assumptions. This issue has sparked significant interest in conditional independence testing over the last several years. We begin by providing an overview of recent advances in conditional independence testing. Afterwards, we discuss how our work addresses limitations in the existing literature. Finally, we motivate our work by reviewing key applications of conditional independence tests for time series in areas such as variable selection and causal discovery.

Artificial Intelligence (AI) models deployed in production frequently face challenges in maintaining their performance in non-stationary environments. This issue is particularly noticeable in medical settings, where temporal dataset shifts often occur. These shifts arise when the distributions of training data differ from those of the data encountered during deployment over time. Further, new labeled data to continuously retrain AI is not typically available in a timely manner due to data access limitations. To address these challenges, we propose a proactive self-adaptive AI approach, or pro-adaptive, where we model the temporal trajectory of AI parameters, allowing us to short-term forecast parameter values. To this end, we use polynomial spline bases, within an extensible Functional Data Analysis framework. We validate our methodology with a logistic regression model addressing prior probability shift, covariate shift, and concept shift. This validation is conducted on both a controlled simulated dataset and a publicly available real-world COVID-19 dataset from Mexico, with various shifts occurring between 2020 and 2024. Our results indicate that this approach enhances the performance of AI against shifts compared to baseline stable models trained at different time distances from the present, without requiring updated training data. This work lays the foundation for pro-adaptive AI research against dynamic, non-stationary environments, being compatible with data protection, in resilient AI production environments for health.

The autonomic nervous system (ANS) is activated during stress, which can have negative effects on cardiovascular health, sleep, the immune system, and mental health. While there are ways to quantify ANS activity in laboratories, there is a paucity of methods that have been validated in real-world contexts. We present the Fitbit Body Response Algorithm, an approach to continuous remote measurement of ANS activation through widely available remote wrist-based sensors. The design was validated via two experiments, a Trier Social Stress Test (n = 45) and ecological momentary assessments (EMA) of perceived stress (n=87), providing both controlled and ecologically valid test data. Model performance predicting perceived stress when using all available sensor modalities was consistent with expectations (accuracy=0.85) and outperformed models with access to only a subset of the signals. We discuss and address challenges to sensing that arise in real world settings that do not present in conventional lab environments.

This study leverages the basic insight that the gradient-flow equation associated with the relative Boltzmann entropy, in relation to a Gaussian reference measure within the Hellinger-Kantorovich (HK) geometry, preserves the class of Gaussian measures. This invariance serves as the foundation for constructing a reduced gradient structure on the parameter space characterizing Gaussian densities. We derive explicit ordinary differential equations that govern the evolution of mean, covariance, and mass under the HK-Boltzmann gradient flow. The reduced structure retains the additive form of the HK metric, facilitating a comprehensive analysis of the dynamics involved. We explore the geodesic convexity of the reduced system, revealing that global convexity is confined to the pure transport scenario, while a variant of sublevel semi-convexity is observed in the general case. Furthermore, we demonstrate exponential convergence to equilibrium through Polyak-Lojasiewicz-type inequalities, applicable both globally and on sublevel sets. By monitoring the evolution of covariance eigenvalues, we refine the decay rates associated with convergence. Additionally, we extend our analysis to non-Gaussian targets exhibiting strong log-lambda-concavity, corroborating our theoretical results with numerical experiments that encompass a Gaussian-target gradient flow and a Bayesian logistic regression application.

In many, if not most, machine learning applications the training data is naturally heterogeneous (e.g. federated learning, adversarial attacks and domain adaptation in neural net training). Data heterogeneity is identified as one of the major challenges in modern day large-scale learning. A classical way to represent heterogeneous data is via a mixture model. In this paper, we study generalization performance and statistical rates when data is sampled from a mixture distribution. We first characterize the heterogeneity of the mixture in terms of the pairwise total variation distance of the sub-population distributions. Thereafter, as a central theme of this paper, we characterize the range where the mixture may be treated as a single (homogeneous) distribution for learning. In particular, we study the generalization performance under the classical PAC framework and the statistical error rates for parametric (linear regression, mixture of hyperplanes) as well as non-parametric (Lipschitz, convex and Hölder-smooth) regression problems. In order to do this, we obtain Rademacher complexity and (local) Gaussian complexity bounds with mixture data, and apply them to get the generalization and convergence rates respectively. We observe that as the (regression) function classes get more complex, the requirement on the pairwise total variation distance gets stringent, which matches our intuition. We also do a finer analysis for the case of mixed linear regression and provide a tight bound on the generalization error in terms of heterogeneity.

Financial data analysis plays a pivotal role in today's business landscape [1, 2, 3, 4, 5, 6, 7], including credit risk assessment (such as loan prediction and credit scoring), fraud detection, and cost optimization, etc. However, when we develop solutions to address financial problems, we will inevitably encounter a number of key challenges [1, 2, 3, 4, 5]. For example, financial data is often voluminous, dynamically and frequently generated in real time, and distributed across diverse locations, making it challenging to process and analyze in a centralized manner[1], e.g., the New Y ork Stock Exchange (NYSE) alone has billions of transactions per day. Similarly, other major exchanges, such as the Shanghai Stock Exchange (SSE) and the London Stock Exchange (LSE), also generate vast amounts of stock data. Additionally, noise and missing values unavoidably occur in financial data, which can cause results and predictions to be skewed (or even completely wrong). These challenges require firms to come up with more efficient and smarter solutions. In recent decades, machine learning has achieved remarkable success across various domains [8, 9, 10], owing to its effective generalization ability and adaptability, and has also received increasing attention in financial data analysis [11, 12], such as credit risk assessment, resource allocation, and cost optimization. However, these classical (supervised) machine learning based solutions, such as logistic regression and random forest, usually implicitly assume that 1) all the data is stored and centralized at one location, typically a single machine, and that we have full access to the entire data; 2) these algorithms expect to run on a single machine with minimal concerns for memory or disk storage limitations; and 3) the provided data is clean and free from outliers introduced by malicious adversaries, as it is stored at a single location equipped with high security protection mechanisms to prevent data corruption. Nonetheless, these assumptions do not always hold in practice.

The machine learning pipeline typically involves the iterative process of (1) collecting the data, (2) preparing the data, (3) learning a model, and (4) evaluating a model. Practitioners recognize the importance of the data preparation phase in terms of its impact on the ability to learn accurate models. In this regard, significant attention is often paid to manipulating the feature set (e.g., selection, transformations, dimensionality reduction). A point that is less well appreciated is that transformations on the target variable can also have a large impact on whether it is possible to learn a suitable model. These transformations may include accounting for subject-specific biases (e.g., in how someone uses a rating scale), contexts (e.g., population size effects), and general trends (e.g., inflation). However, this point has received a much more cursory treatment in the existing literature. The goal of this paper is three-fold. First, we aim to highlight the importance of this problem by showing when transforming the target variable has been useful in practice. Second, we will provide a set of generic ``rules of thumb'' that indicate situations when transforming the target variable may be needed. Third, we will discuss which transformations should be considered in a given situation.

Language models can behave in unexpected and unsafe ways, and so it is valuable to monitor their outputs. Internal activations of language models encode additional information that could be useful for this. The baseline approach for activation monitoring is some variation of linear probing on a particular layer: starting from a labeled dataset, train a logistic regression classifier on that layer's activations. Recent work has proposed several approaches which may improve on naive linear probing, by leveraging additional computation. One class of techniques, which we call "prompted probing," leverages test time computation to improve monitoring by (1) prompting the model with a description of the monitoring task, and (2) applying a learned linear probe to resulting activations. Another class of techniques uses computation at train time: training sparse autoencoders offline to identify an interpretable basis for the activations, and e.g. max-pooling activations across tokens using that basis before applying a linear probe. However, one can also prompt the model with a description of the monitoring task and use its output directly. We develop and test novel refinements of these methods and compare them against each other. We find asking the model zero-shot is a reasonable baseline when inference-time compute is not limited; however, activation probing methods can substantially outperform this baseline given sufficient training data. Specifically, we recommend prompted probing when inference-time compute is available, due to its superior data efficiency and good generalization performance. Alternatively, if inference-time compute is limited, we find SAE-based probing methods outperform raw activation probing.