This paper demonstrates our vision for knowledge graphs that assist machines to reason about the cause of signals observed by sensors. We show how the approach allows for constructing smarter surveillance systems that reason about the most likely cause (e.g., an attacker breaking a window) of a signal rather than acting directly on the received signal without consideration for how it was produced.

Prediction of individual's race and ethnicity plays an important role in social science and public health research. Examples include studies of racial disparity in health and voting. Recently, Bayesian Improved Surname Geocoding (BISG), which uses Bayes' rule to combine information from Census surname files with the geocoding of an individual's residence, has emerged as a leading methodology for this prediction task. Unfortunately, BISG suffers from two Census data problems that contribute to unsatisfactory predictive performance for minorities. First, the decennial Census often contains zero counts for minority racial groups in the Census blocks where some members of those groups reside. Second, because the Census surname files only include frequent names, many surnames -- especially those of minorities -- are missing from the list. To address the zero counts problem, we introduce a fully Bayesian Improved Surname Geocoding (fBISG) methodology that accounts for potential measurement error in Census counts by extending the naive Bayesian inference of the BISG methodology to full posterior inference. To address the missing surname problem, we supplement the Census surname data with additional data on last, first, and middle names taken from the voter files of six Southern states where self-reported race is available. Our empirical validation shows that the fBISG methodology and name supplements significantly improve the accuracy of race imputation across all racial groups, and especially for Asians. The proposed methodology, together with additional name data, is available via the open-source software WRU.

This paper introduces Probabilistic Deduction (PD) as an approach to probabilistic structured argumentation. A PD framework is composed of probabilistic rules (p-rules). As rules in classical structured argumentation frameworks, p-rules form deduction systems. In addition, p-rules also represent conditional probabilities that define joint probability distributions. With PD frameworks, one performs probabilistic reasoning by solving Rule-Probabilistic Satisfiability. At the same time, one can obtain an argumentative reading to the probabilistic reasoning with arguments and attacks. In this work, we introduce a probabilistic version of the Closed-World Assumption (P-CWA) and prove that our probabilistic approach coincides with the complete extension in classical argumentation under P-CWA and with maximum entropy reasoning. We present several approaches to compute the joint probability distribution from p-rules for achieving a practical proof theory for PD. PD provides a framework to unify probabilistic reasoning with argumentative reasoning. This is the first work in probabilistic structured argumentation where the joint distribution is not assumed form external sources.

Probabilistic programs provide an expressive representation language for generative models. Given a probabilistic program, we are interested in the task of posterior inference: estimating a latent variable given a set of observed variables. Existing techniques for inference in probabilistic programs often require choosing many hyper-parameters, are computationally expensive, and/or only work for restricted classes of programs. Here we formulate inference as masked language modeling: given a program, we generate a supervised dataset of variables and assignments, and randomly mask a subset of the assignments. We then train a neural network to unmask the random values, defining an approximate posterior distribution. By optimizing a single neural network across a range of programs we amortize the cost of training, yielding a "foundation" posterior able to do zero-shot inference for new programs. The foundation posterior can also be fine-tuned for a particular program and dataset by optimizing a variational inference objective. We show the efficacy of the approach, zero-shot and fine-tuned, on a benchmark of STAN programs.

Road accidents have become the eight leading cause of death all over the world. Lots of these accidents are due to a driver's inattention or lack of focus, due to fatigue. Various factors cause driver's fatigue. This paper considers all the measureable data that manifest driver's fatigue, namely those manifested in the vehicle measureable data while driving as well as the driver's physical and physiological data. Each of the three main factors are further subdivided into smaller details. For example, the vehicle's data is composed of the values obtained from the steering wheel's angle, yaw angle, the position on the lane, and the speed and acceleration of the vehicle while moving. Ontological knowledge and rules for driver fatigue detection are to be integrated into an intelligent system so that on the first sign of dangerous level of fatigue is detected, a warning notification is sent to the driver. This work is intended to contribute to safe road driving.

EHR repositories contain large amounts of wide-ranging patient and treatment information and are essential for the development of individualized treatments in the context of disease progression [14]. With the broad adoption of EHR in the US, a large variety of data types are now routinely collected over long periods of time. This has ushered in an era of research focused on the applications and development of data-analytic tools for mining historical records of medical data to drive novel insight. Broadly, the extraction of meaningful patterns through unsupervised learning [31, 23, 17, 10] and the prediction of outcomes through supervised learning [22, 43, 41, 40, 37, 16, 13, 34, 3] are two important directions. Unsupervised methods can be applied towards many different tasks, such as prediction, imputation, and simulation; and often contain a model of the underlying structure in the data [29]. This underlying structure is not directly observed and can lead to insights that are otherwise difficult to produce, especially for large and complex data sets.

Diffusion-based generative models have achieved promising results recently, but raise an array of open questions in terms of conceptual understanding, theoretical analysis, algorithm improvement and extensions to discrete, structured, non-Euclidean domains. This work tries to re-exam the overall framework, in order to gain better theoretical understandings and develop algorithmic extensions for data from arbitrary domains. By viewing diffusion models as latent variable models with unobserved diffusion trajectories and applying maximum likelihood estimation (MLE) with latent trajectories imputed from an auxiliary distribution, we show that both the model construction and the imputation of latent trajectories amount to constructing diffusion bridge processes that achieve deterministic values and constraints at end point, for which we provide a systematic study and a suit of tools. Leveraging our framework, we present 1) a first theoretical error analysis for learning diffusion generation models, and 2) a simple and unified approach to learning on data from different discrete and constrained domains. Experiments show that our methods perform superbly on generating images, semantic segments and 3D point clouds.

In many scenarios such as genome-wide association studies where dependences between variables commonly exist, it is often of interest to infer the interaction effects in the model. However, testing pairwise interactions among millions of variables in complex and high-dimensional data suffers from low statistical power and huge computational cost. To address these challenges, we propose a two-stage testing procedure with false discovery rate (FDR) control, which is known as a less conservative multiple-testing correction. Theoretically, the difficulty in the FDR control dues to the data dependence among test statistics in two stages, and the fact that the number of hypothesis tests conducted in the second stage depends on the screening result in the first stage. By using the Cram\'er type moderate deviation technique, we show that our procedure controls FDR at the desired level asymptotically in the generalized linear model (GLM), where the model is allowed to be misspecified. In addition, the asymptotic power of the FDR control procedure is rigorously established. We demonstrate via comprehensive simulation studies that our two-stage procedure is computationally more efficient than the classical BH procedure, with a comparable or improved statistical power. Finally, we apply the proposed method to a bladder cancer data from dbGaP where the scientific goal is to identify genetic susceptibility loci for bladder cancer.

This paper attempts to establish the theoretical foundation for the emerging super-model paradigm via domain adaptation, where one first trains a very large-scale model, {\it i.e.}, super model (or foundation model in some other papers), on a large amount of data and then adapts it to various specific domains. Super-model paradigms help reduce computational and data cost and carbon emission, which is critical to AI industry, especially enormous small and medium-sized enterprises. We model the super-model paradigm as a two-stage diffusion process: (1) in the pre-training stage, the model parameter diffuses from random initials and converges to a steady distribution; and (2) in the fine-tuning stage, the model parameter is transported to another steady distribution. Both training stages can be mathematically modeled by the Uhlenbeck-Ornstein process which converges to two Maxwell-Boltzmann distributions, respectively, each of which characterizes the corresponding convergent model. An $\mathcal O(1/\sqrt{N})$ generalization bound is then established via PAC-Bayesian framework. The theory finds that the generalization error of the fine-tuning stage is dominant in domain adaptation. In addition, our theory suggests that the generalization is determined by a new measure that characterizes the domain discrepancy between the source domain and target domain, based on the covariance matrices and the shift of the converged local minimum.

However, estimating treatment effects using EHR data poses several challenges, including time-varying confounding, repeated and temporally non-aligned measurements of covariates, treatment assignments and outcomes, and loss-to-follow-up due to dropout. Here, we develop the Subgroup Discovery for Longitudinal Data (SDLD) algorithm, a tree-based algorithm for discovering subgroups with heterogeneous treatment effects using longitudinal data by combining the generalized interaction tree algorithm, a general data-driven method for subgroup discovery, with longitudinal targeted maximum likelihood estimation. We apply the algorithm to EHR data to discover subgroups of people living with human immunodeficiency virus (HIV) who are at higher risk of weight gain when receiving dolutegravir-containing antiretroviral therapies (ARTs) versus when receiving non dolutegravir-containing ARTs. Key words: Causal Inference; Dolutegravir; Electronic health record; Heterogeneity of treatment effects; Longitudinal targeted maximum likelihood estimation; Machine learning; Recursive partitioning; Subgroup discovery.