Search

Speech emotion recognition (SER) systems often struggle in real-world environments, where ambient noise severely degrades their performance. This paper explores a novel approach that exploits prior knowledge of testing environments to maximize SER performance under noisy conditions. To address this task, we propose a text-guided, environment-aware training where an SER model is trained with contaminated speech samples and their paired noise description. We use a pre-trained text encoder to extract the text-based environment embedding and then fuse it to a transformer-based SER model during training and inference. We demonstrate the effectiveness of our approach through our experiment with the MSP-Podcast corpus and real-world additive noise samples collected from the Freesound repository. Our experiment indicates that the text-based environment descriptions processed by a large language model (LLM) produce representations that improve the noise-robustness of the SER system. In addition, our proposed approach with an LLM yields better performance than our environment-agnostic baselines, especially in low signal-to-noise ratio (SNR) conditions. When testing at -5dB SNR level, our proposed method shows better performance than our best baseline model by 31.8 % (arousal), 23.5% (dominance), and 9.5% (valence).

Mobile technology (mobile phones and wearable devices) generates continuous data streams encompassing outcomes, exposures and covariates, presented as intensive longitudinal or multivariate time series data. The high frequency of measurements enables granular and dynamic evaluation of treatment effect, revealing their persistence and accumulation over time. Existing methods predominantly focus on the contemporaneous effect, temporal-average, or population-average effects, assuming stationarity or invariance of treatment effects over time, which are inadequate both conceptually and statistically to capture dynamic treatment effects in personalized mobile health data. We here propose new causal estimands for multivariate time series in N-of-1 studies. These estimands summarize how time-varying exposures impact outcomes in both short- and long-term. We propose identifiability assumptions and a g-formula estimator that accounts for exposure-outcome and outcome-covariate feedback. The g-formula employs a state space model framework innovatively to accommodate time-varying behavior of treatment effects in non-stationary time series. We apply the proposed method to a multi-year smartphone observational study of bipolar patients and estimate the dynamic effect of phone-based communication on mood of patients with bipolar disorder in an N-of-1 setting. Our approach reveals substantial heterogeneity in treatment effects over time and across individuals. A simulation-based strategy is also proposed for the development of a short-term, dynamic, and personalized treatment recommendation based on patient's past information, in combination with a novel positivity diagnostics plot, validating proper causal inference in time series data.

We introduce a family of parsimonious network models that are intended to generalize the configuration model to temporal settings. We present consistent estimators for the model parameters and perform numerical simulations to illustrate the properties of the estimators on finite samples. We also develop analytical solutions for basic and effective reproductive numbers for the early stage of discrete-time SIR spreading process. We apply three distinct temporal configuration models to empirical student proximity networks and compare their performance., Comment: 22 pages and 2 figures

Depth measures have gained popularity in the statistical literature for defining level sets in complex data structures like multivariate data, functional data, and graphs. Despite their versatility, integrating depth measures into regression modeling for establishing prediction regions remains underexplored. To address this gap, we propose a novel method utilizing a model-free uncertainty quantification algorithm based on conditional depth measures and conditional kernel mean embeddings. This enables the creation of tailored prediction and tolerance regions in regression models handling complex statistical responses and predictors in separable Hilbert spaces. Our focus in this paper is exclusively on examples where the response is a functional data object. To enhance practicality, we introduce a conformal prediction algorithm, providing non-asymptotic guarantees in the derived prediction region. Additionally, we establish both conditional and unconditional consistency results and fast convergence rates in some special homoscedastic cases. We evaluate the model finite sample performance in extensive simulation studies with different function objects as probability distributions and functional data. Finally, we apply the approach in a digital health application related to physical activity, aiming to offer personalized recommendations in the US. population based on individuals' characteristics.

When modeling the dynamics of infectious disease, the incorporation of contact network information allows for the capture of the non-randomness and heterogeneity of realistic contact patterns. Oftentimes, it is assumed that the underlying contact pattern is known with perfect certainty. However, in realistic settings, the observed data often serves as an imperfect proxy of the actual contact patterns in the population. Furthermore, the epidemic in the real world are often not fully observed; event times such as infection and recovery times may be missing. In order to conduct accurate inferences on parameters of contagion spread, it is crucial to incorporate these sources of uncertainty. In this paper, we propose the use of Mixture Density Network compressed ABC (MDN-ABC) to learn informative summary statistics for the available data. This method will allow for Bayesian inference on the epidemic parameters of a contagious process, while accounting for imperfect observations on the epidemic and the contact network. We will demonstrate the use of this method on simulated epidemics and networks, and extend this framework to analyze the spread of Tattoo Skin Disease (TSD) among bottlenose dolphins in Shark Bay, Australia., Comment: 27 pages, 7 figures

Complex survey designs are commonly employed in many medical cohorts. In such scenarios, developing case-specific predictive risk score models that reflect the unique characteristics of the study design is essential. This approach is key to minimizing potential selective biases in results. The objectives of this paper are: (i) To propose a general predictive framework for regression and classification using neural network (NN) modeling, which incorporates survey weights into the estimation process; (ii) To introduce an uncertainty quantification algorithm for model prediction, tailored for data from complex survey designs; (iii) To apply this method in developing robust risk score models to assess the risk of Diabetes Mellitus in the US population, utilizing data from the NHANES 2011-2014 cohort. The theoretical properties of our estimators are designed to ensure minimal bias and the statistical consistency, thereby ensuring that our models yield reliable predictions and contribute novel scientific insights in diabetes research. While focused on diabetes, this NN predictive framework is adaptable to create clinical models for a diverse range of diseases and medical cohorts. The software and the data used in this paper is publicly available on GitHub.

In this paper, we introduce a kNN-based regression method that synergizes the scalability and adaptability of traditional non-parametric kNN models with a novel variable selection technique. This method focuses on accurately estimating the conditional mean and variance of random response variables, thereby effectively characterizing conditional distributions across diverse scenarios.Our approach incorporates a robust uncertainty quantification mechanism, leveraging our prior estimation work on conditional mean and variance. The employment of kNN ensures scalable computational efficiency in predicting intervals and statistical accuracy in line with optimal non-parametric rates. Additionally, we introduce a new kNN semi-parametric algorithm for estimating ROC curves, accounting for covariates. For selecting the smoothing parameter k, we propose an algorithm with theoretical guarantees.Incorporation of variable selection enhances the performance of the method significantly over conventional kNN techniques in various modeling tasks. We validate the approach through simulations in low, moderate, and high-dimensional covariate spaces. The algorithm's effectiveness is particularly notable in biomedical applications as demonstrated in two case studies. Concluding with a theoretical analysis, we highlight the consistency and convergence rate of our method over traditional kNN models, particularly when the underlying regression model takes values in a low-dimensional space.

Network models are increasingly used to study infectious disease spread. Exponential Random Graph models have a history in this area, with scalable inference methods now available. An alternative approach uses mechanistic network models. Mechanistic network models directly capture individual behaviors, making them suitable for studying sexually transmitted diseases. Combining mechanistic models with Approximate Bayesian Computation allows flexible modeling using domain-specific interaction rules among agents, avoiding network model oversimplifications. These models are ideal for longitudinal settings as they explicitly incorporate network evolution over time. We implemented a discrete-time version of a previously published continuous-time model of evolving contact networks for men who have sex with men (MSM) and proposed an ABC-based approximate inference scheme for it. As expected, we found that a two-wave longitudinal study design improves the accuracy of inference compared to a cross-sectional design. However, the gains in precision in collecting data twice, up to 18%, depend on the spacing of the two waves and are sensitive to the choice of summary statistics. In addition to methodological developments, our results inform the design of future longitudinal network studies in sexually transmitted diseases, specifically in terms of what data to collect from participants and when to do so., Comment: 27 pages, Figure 7

Network science explores intricate connections among objects, employed in diverse domains like social interactions, fraud detection, and disease spread. Visualization of networks facilitates conceptualizing research questions and forming scientific hypotheses. Networks, as mathematical high-dimensional objects, require dimensionality reduction for (planar) visualization. Visualizing empirical networks present additional challenges. They often contain false positive (spurious) and false negative (missing) edges. Traditional visualization methods don't account for errors in observation, potentially biasing interpretations. Moreover, contemporary network data includes rich nodal attributes. However, traditional methods neglect these attributes when computing node locations. Our visualization approach aims to leverage nodal attribute richness to compensate for network data limitations. We employ a statistical model estimating the probability of edge connections between nodes based on their covariates. We enhance the Fruchterman-Reingold algorithm to incorporate estimated dyad connection probabilities, allowing practitioners to balance reliance on observed versus estimated edges. We explore optimal smoothing levels, offering a natural way to include relevant nodal information in layouts. Results demonstrate the effectiveness of our method in achieving robust network visualization, providing insights for improved analysis., Comment: 23 pages, 6 figures

Individual-based models of contagious processes are useful for predicting epidemic trajectories and informing intervention strategies. In such models, the incorporation of contact network information can capture the non-randomness and heterogeneity of realistic contact dynamics. In this paper, we consider Bayesian inference on the spreading parameters of an SIR contagion on a known, static network, where information regarding individual disease status is known only from a series of tests (positive or negative disease status). When the contagion model is complex or information such as infection and removal times is missing, the posterior distribution can be difficult to sample from. Previous work has considered the use of Approximate Bayesian Computation (ABC), which allows for simulation-based Bayesian inference on complex models. However, ABC methods usually require the user to select reasonable summary statistics. Here, we consider an inference scheme based on the Mixture Density Network compressed ABC (MDN-ABC), which minimizes the expected posterior entropy in order to learn informative summary statistics. This allows us to conduct Bayesian inference on the parameters of a partially observed contagious process while also circumventing the need for manual summary statistic selection. This methodology can be extended to incorporate additional simulation complexities, including behavioral change after positive tests or false test results., Comment: 27 pages, 7 figures

We propose a nonparametric additive model for estimating interpretable value functions in reinforcement learning. Learning effective adaptive clinical interventions that rely on digital phenotyping features is a major for concern medical practitioners. With respect to spine surgery, different post-operative recovery recommendations concerning patient mobilization can lead to significant variation in patient recovery. While reinforcement learning has achieved widespread success in domains such as games, recent methods heavily rely on black-box methods, such neural networks. Unfortunately, these methods hinder the ability of examining the contribution each feature makes in producing the final suggested decision. While such interpretations are easily provided in classical algorithms such as Least Squares Policy Iteration, basic linearity assumptions prevent learning higher-order flexible interactions between features. In this paper, we present a novel method that offers a flexible technique for estimating action-value functions without making explicit parametric assumptions regarding their additive functional form. This nonparametric estimation strategy relies on incorporating local kernel regression and basis expansion to obtain a sparse, additive representation of the action-value function. Under this approach, we are able to locally approximate the action-value function and retrieve the nonlinear, independent contribution of select features as well as joint feature pairs. We validate the proposed approach with a simulation study, and, in an application to spine disease, uncover recovery recommendations that are inline with related clinical knowledge., Comment: 28 pages, 13 figures

Gaussian processes (GPs) are sophisticated distributions to model functional data. Whilst theoretically appealing, they are computationally cumbersome except for small datasets. We implement two methods for scaling GP inference in Stan: First, a general sparse approximation using a directed acyclic dependency graph; second, a fast, exact method for regularly spaced data modeled by GPs with stationary kernels using the fast Fourier transform. Based on benchmark experiments, we offer guidance for practitioners to decide between different methods and parameterizations. We consider two real-world examples to illustrate the package. The implementation follows Stan's design and exposes performant inference through a familiar interface. Full posterior inference for ten thousand data points is feasible on a laptop in less than 20 seconds. Details on how to get started using the popular interfaces cmdstanpy for Python and cmdstanr for R are provided., Comment: 32 pages, 6 figures

Regression models applied to network data where node attributes are the dependent variables poses a methodological challenge. As has been well studied, naive regression neither properly accounts for community structure, nor does it account for the dependent variable acting as both model outcome and covariate. To address this methodological gap, we propose a network regression model motivated by the important observation that controlling for community structure can, when a network is modular, significantly account for meaningful correlation between observations induced by network connections. We propose a generalized estimating equation (GEE) approach to learn model parameters based on clusters defined through any single-membership community detection algorithm applied to the observed network. We provide a necessary condition on the network size and edge formation probabilities to establish the asymptotic normality of the model parameters under the assumption that the graph structure is a stochastic block model. We evaluate the performance of our approach through simulations and apply it to estimate the joint impact of baseline covariates and network effects on COVID-19 incidence rate among countries connected by a network of commercial airline traffic. We find that during the beginning of the pandemic the network effect has some influence, the percentage of urban population has more influence on the incidence rate compared to the network effect after the travel ban was in effect., Comment: 24 pages, 2 figures, 3 tables

The use of digital devices to collect data in mobile health (mHealth) studies introduces a novel application of time series methods, with the constraint of potential data missing at random (MAR) or missing not at random (MNAR). In time series analysis, testing for stationarity is an important preliminary step to inform appropriate later analyses. The augmented Dickey-Fuller (ADF) test was developed to test the null hypothesis of unit root non-stationarity, under no missing data. Beyond recommendations under data missing completely at random (MCAR) for complete case analysis or last observation carry forward imputation, researchers have not extended unit root non-stationarity testing to a context with more complex missing data mechanisms. Multiple imputation with chained equations, Kalman smoothing imputation, and linear interpolation have also been proposed for time series data, however such methods impose constraints on the autocorrelation structure, and thus impact unit root testing. We propose maximum likelihood estimation and multiple imputation using state space model approaches to adapt the ADF test to a context with missing data. We further develop sensitivity analysis techniques to examine the impact of MNAR data. We evaluate the performance of existing and proposed methods across different missing mechanisms in extensive simulations and in their application to a multi-year smartphone study of bipolar patients.

The ubiquity of personal digital devices offers unprecedented opportunities to study human behavior. Current state-of-the-art methods quantify physical activity using 'activity counts,' a measure which overlooks specific types of physical activities. We proposed a walking recognition method for sub-second tri-axial accelerometer data, in which activity classification is based on the inherent features of walking: intensity, periodicity, and duration. We validated our method against 20 publicly available, annotated datasets on walking activity data collected at various body locations (thigh, waist, chest, arm, wrist). We demonstrated that our method can estimate walking periods with high sensitivity and specificity: average sensitivity ranged between 0.92 and 0.97 across various body locations, and average specificity for common daily activities was typically above 0.95. We also assessed the method's algorithmic fairness to demographic and anthropometric variables and measurement contexts (body location, environment). Finally, we have released our method as open-source software in MATLAB and Python., Comment: 39 pages, 4 figures (incl. 1 supplementary), and 5 tables (incl. 2 supplementary)

Mobile technology enables unprecedented continuous monitoring of an individual's behavior, social interactions, symptoms, and other health conditions, presenting an enormous opportunity for therapeutic advancements and scientific discoveries regarding the etiology of psychiatric illness. Continuous collection of mobile data results in the generation of a new type of data: entangled multivariate time series of outcome, exposure, and covariates. Missing data is a pervasive problem in biomedical and social science research, and the Ecological Momentary Assessment (EMA) using mobile devices in psychiatric research is no exception. However, the complex structure of multivariate time series introduces new challenges in handling missing data for proper causal inference. Data imputation is commonly recommended to enhance data utility and estimation efficiency. The majority of available imputation methods are either designed for longitudinal data with limited follow-up times or for stationary time series, which are incompatible with potentially non-stationary time series. In the field of psychiatry, non-stationary data are frequently encountered as symptoms and treatment regimens may experience dramatic changes over time. To address missing data in possibly non-stationary multivariate time series, we propose a novel multiple imputation strategy based on the state space model (SSMmp) and a more computationally efficient variant (SSMimpute). We demonstrate their advantages over other widely used missing data strategies by evaluating their theoretical properties and empirical performance in simulations of both stationary and non-stationary time series, subject to various missing mechanisms. We apply the SSMimpute to investigate the association between social network size and negative mood using a multi-year observational smartphone study of bipolar patients, controlling for confounding variables.

Extracting low-dimensional summary statistics from large datasets is essential for efficient (likelihood-free) inference. We characterise different classes of summaries and demonstrate their importance for correctly analysing dimensionality reduction algorithms. We propose obtaining summaries by minimising the expected posterior entropy (EPE) under the prior predictive distribution of the model. Many existing methods are equivalent to or are special or limiting cases of minimising the EPE. We develop a method to obtain high-fidelity summaries that minimise the EPE; we apply it to benchmark and real-world examples. We both offer a unifying perspective for obtaining informative summaries and provide concrete recommendations for practitioners., Comment: 32 pages, 6 figures, 1 table

Physical activity patterns can be informative about a patient's health status. Traditionally, activity data have been gathered using patient self-report. However, these subjective data can suffer from bias and are difficult to collect over long time periods. Smartphones offer an opportunity to address these challenges. The smartphone has built-in sensors that can be programmed to collect data objectively, unobtrusively, and continuously. Due to their widespread adoption, smartphones are also accessible to most of the population. A main challenge in smartphone-based activity recognition is extracting information optimally from multiple sensors to identify the unique features of different activities. In our study, we analyze data collected by the accelerometer and gyroscope, which measure the phone's acceleration and angular velocity, respectively. We propose an extension to the "movelet method" that jointly incorporates both sensors. We also apply this joint-sensor method to a data set we collected previously. The findings show that combining data from the two sensors can result in more accurate activity recognition than using each sensor alone. For example, the joint-sensor method reduces errors of the gyroscope-only method in differentiating between standing and sitting. It also reduces errors of the accelerometer-only method in classifying vigorous activities.

Mechanistic network models specify the mechanisms by which networks grow and change, allowing researchers to investigate complex systems using both simulation and analytical techniques. Unfortunately, it is difficult to write likelihoods for instances of graphs generated with mechanistic models because of a combinatorial explosion in outcomes of repeated applications of the mechanism. Thus it is near impossible to estimate the parameters using maximum likelihood estimation. In this paper, we propose treating node sequence in a growing network model as an additional parameter, or as a missing random variable, and maximizing over the resulting likelihood. We develop this framework in the context of a simple mechanistic network model, used to study gene duplication and divergence, and test a variety of algorithms for maximizing the likelihood in simulated graphs. We also run the best-performing algorithm on a human protein-protein interaction network and four non-human protein-protein interaction networks. Although we focus on a specific mechanistic network model here, the proposed framework is more generally applicable to reversible models.

Minimum spanning trees (MSTs) are used in a variety of fields, from computer science to geography. Infectious disease researchers have used them to infer the transmission pathway of certain pathogens. However, these are often the MSTs of sample networks, not population networks, and surprisingly little is known about what can be inferred about a population MST from a sample MST. We prove that if $n$ nodes (the sample) are selected uniformly at random from a complete graph with $N$ nodes and unique edge weights (the population), the probability that an edge is in the population graph's MST given that it is in the sample graph's MST is $\frac{n}{N}$. We use simulation to investigate this conditional probability for $G(N,p)$ graphs, Barab\'{a}si-Albert (BA) graphs, graphs whose nodes are distributed in $\mathbb{R}^2$ according to a bivariate standard normal distribution, and an empirical HIV genetic distance network. Broadly, results for the complete, $G(N,p)$, and normal graphs are similar, and results for the BA and empirical HIV graphs are similar. We recommend that researchers use an edge-weighted random walk to sample nodes from the population so that they maximize the probability that an edge is in the population MST given that it is in the sample MST., Comment: 26 pages, 6 figures

Selecting a small set of informative features from a large number of possibly noisy candidates is a challenging problem with many applications in machine learning and approximate Bayesian computation. In practice, the cost of computing informative features also needs to be considered. This is particularly important for networks because the computational costs of individual features can span several orders of magnitude. We addressed this issue for the network model selection problem using two approaches. First, we adapted nine feature selection methods to account for the cost of features. We show for two classes of network models that the cost can be reduced by two orders of magnitude without considerably affecting classification accuracy (proportion of correctly identified models). Second, we selected features using pilot simulations with smaller networks. This approach reduced the computational cost by a factor of 50 without affecting classification accuracy. To demonstrate the utility of our approach, we applied it to three different yeast protein interaction networks and identified the best-fitting duplication divergence model., Comment: 34 pages, 6 figures

Approximate Bayesian computation (ABC) is a simulation-based likelihood-free method applicable to both model selection and parameter estimation. ABC parameter estimation requires the ability to forward simulate datasets from a candidate model, but because the sizes of the observed and simulated datasets usually need to match, this can be computationally expensive. Additionally, since ABC inference is based on comparisons of summary statistics computed on the observed and simulated data, using computationally expensive summary statistics can lead to further losses in efficiency. ABC has recently been applied to the family of mechanistic network models, an area that has traditionally lacked tools for inference and model choice. Mechanistic models of network growth repeatedly add nodes to a network until it reaches the size of the observed network, which may be of the order of millions of nodes. With ABC, this process can quickly become computationally prohibitive due to the resource intensive nature of network simulations and evaluation of summary statistics. We propose two methodological developments to enable the use of ABC for inference in models for large growing networks. First, to save time needed for forward simulating model realizations, we propose a procedure to extrapolate (via both least squares and Gaussian processes) summary statistics from small to large networks. Second, to reduce computation time for evaluating summary statistics, we use sample-based rather than census-based summary statistics. We show that the ABC posterior obtained through this approach, which adds two additional layers of approximation to the standard ABC, is similar to a classic ABC posterior. Although we deal with growing network models, both extrapolated summaries and sampled summaries are expected to be relevant in other ABC settings where the data are generated incrementally., Comment: 28 pages, 16 figures