Search

In the dynamic hospital setting, decision support can be a valuable tool for improving patient outcomes. Data-driven inference of future outcomes is challenging in this dynamic setting, where long sequences such as laboratory tests and medications are updated frequently. This is due in part to heterogeneity of data types and mixed-sequence types contained in variable length sequences. In this work we design a probabilistic unsupervised model for multiple arbitrary-length sequences contained in hospitalization Electronic Health Record (EHR) data. The model uses a latent variable structure and captures complex relationships between medications, diagnoses, laboratory tests, neurological assessments, and medications. It can be trained on original data, without requiring any lossy transformations or time binning. Inference algorithms are derived that use partial data to infer properties of the complete sequences, including their length and presence of specific values. We train this model on data from subjects receiving medical care in the Kaiser Permanente Northern California integrated healthcare delivery system. The results are evaluated against held-out data for predicting the length of sequences and presence of Intensive Care Unit (ICU) in hospitalization bed sequences. Our method outperforms a baseline approach, showing that in these experiments the trained model captures information in the sequences that is informative of their future values.

We develop an unsupervised probabilistic model for heterogeneous Electronic Health Record (EHR) data. Utilizing a mixture model formulation, our approach directly models sequences of arbitrary length, such as medications and laboratory results. This allows for subgrouping and incorporation of the dynamics underlying heterogeneous data types. The model consists of a layered set of latent variables that encode underlying structure in the data. These variables represent subject subgroups at the top layer, and unobserved states for sequences in the second layer. We train this model on episodic data from subjects receiving medical care in the Kaiser Permanente Northern California integrated healthcare delivery system. The resulting properties of the trained model generate novel insight from these complex and multifaceted data. In addition, we show how the model can be used to analyze sequences that contribute to assessment of mortality likelihood.

Functional magnetic resonance imaging (fMRI) is a neuroimaging modality that captures the blood oxygen level in a subject's brain while the subject either rests or performs a variety of functional tasks under different conditions. Given fMRI data, the problem of inferring the task, known as task state decoding, is challenging due to the high dimensionality (hundreds of million sampling points per datum) and complex spatio-temporal blood flow patterns inherent in the data. In this work, we propose to tackle the fMRI task state decoding problem by casting it as a 4D spatio-temporal classification problem. We present a novel architecture called Brain Attend and Decode (BAnD), that uses residual convolutional neural networks for spatial feature extraction and self-attention mechanisms for temporal modeling. We achieve significant performance gain compared to previous works on a 7-task benchmark from the large-scale Human Connectome Project-Young Adult (HCP-YA) dataset. We also investigate the transferability of BAnD's extracted features on unseen HCP tasks, either by freezing the spatial feature extraction layers and retraining the temporal model, or finetuning the entire model. The pre-trained features from BAnD are useful on similar tasks while finetuning them yields competitive results on unseen tasks/conditions.

After Pareto distribution has been validated for sea clutter returns in varied scenarios, some heuristics of adaptive-thresholding appeared in the literature for constant false alarm rate (CFAR) criteria. These schemes used the same adaptive-thresholding form that was originally derived for detecting Swerling-I (exponential) target in exponentially distributed clutter. Statistical procedures obtained under such idealistic assumptions would affect the detection performance when applied to newer target and clutter models, esp. heavy tail distributions like Pareto. Further, in addition to the sea clutter returns, it has also been reported that Generalized Pareto distribution fits best for the measured Radar-cross-section (RCS) data of a SAAB aircraft. Therefore, in Radar application scenarios like Airborne Warning and Control System (AWACS), when both the target and clutter are Pareto distributed, we pose the detection problem as a two-sample, Pareto vs. Pareto composite hypothesis testing problem. We address this problem by corroborating the binary hypothesis framework instead of the conventional way of tweaking the existing adaptive-thresholding CFAR detector. Whereby, for the composite case, considering no knowledge of both scale and shape parameters of Pareto distributed clutter, we derive the new adaptive-thresholding detector based on the generalized likelihood ratio test (GLRT) statistic. We further show that our proposed adaptive-thresholding detector has a CFAR property. We provide extensive simulation results to demonstrate the performance of the proposed detector., Comment: 21 pages, 5 figures

We propose multivariate nonstationary Gaussian processes for jointly modeling multiple clinical variables, where the key parameters, length-scales, standard deviations and the correlations between the observed output, are all time dependent. We perform posterior inference via Hamiltonian Monte Carlo (HMC). We also provide methods for obtaining computationally efficient gradient-based maximum a posteriori (MAP) estimates. We validate our model on synthetic data as well as on electronic health records (EHR) data from Kaiser Permanente (KP). We show that the proposed model provides better predictive performance over a stationary model as well as uncovers interesting latent correlation processes across vitals which are potentially predictive of patient risk.

Gaussian processes are a flexible Bayesian nonparametric modelling approach that has been widely applied but poses computational challenges. To address the poor scaling of exact inference methods, approximation methods based on sparse Gaussian processes (SGP) are attractive. An issue faced by SGP, especially in latent variable models, is the inefficient learning of the inducing inputs, which leads to poor model prediction. We propose a regularization approach by balancing the reconstruction performance of data and the approximation performance of the model itself. This regularization improves both inference and prediction performance. We extend this regularization approach into latent variable models with SGPs and show that performing variational inference (VI) on those models is equivalent to performing VI on a related empirical Bayes model.

In this paper, a resource allocation problem for an opportunistic cooperative cognitive radio network is considered, where cognitive radio nodes send their hard decisions to the fusion center. The fusion center plays dual role, i.e., takes the global decision (i.e., decision about the primary user's activity) as well as allocates transmission time durations among cognitive radio nodes. Revenue based utility functions are considered at the fusion center and cognitive radio nodes. An optimization problem is formulated to maximize the fusion center's revenue while satisfying some well defined constraints. User selection among cognitive radio nodes is performed in order to make the optimization problem feasible., Comment: 32 pages, 8 figures, Journal

In this work, we investigate the successful data communication probability of an energy harvesting co-operative cognitive radio network (CRN) in the presence of Poisson field of primary users (PU). We consider the scenario where, after harvesting energy from primary transmitters (PTs), the secondary transmitter (ST) would transmit its symbol towards secondary destination (SD) through a suitable secondary relay from group of randomly scattered idle nodes within a circular region. We have considered several relay selection criteria in our work for a better relay node selection. We have also analytically evaluated the performance of secondary transmitter in terms of probability of successful symbol transmission. The relationship between the performance of ST and several network entities like density of PUs, transmit power of PTs and required transmit power of ST have been investigated through detailed analysis. The non-trivial trade-off between benefit of energy harvesting and interference from PTs has been explored in this present work. Numerical results are provided to verify the precision of derived analytical expressions.

In this paper, we propose a locally optimum detection (LOD) scheme for detecting a weak radioactive source buried in background clutter. We develop a decentralized algorithm, based on alternating direction method of multipliers (ADMM), for implementing the proposed scheme in autonomous sensor networks. Results show that algorithm performance approaches the centralized clairvoyant detection algorithm in the low SNR regime, and exhibits excellent convergence rate and scaling behavior (w.r.t. number of nodes). We also devise a low-overhead, robust ADMM algorithm for Byzantine-resilient detection, and demonstrate its robustness to data falsification attacks., Comment: Proceedings of the 2017 IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP 2017), 10.-13. December 2017, Curacao, Dutch Antilles

Characterizing a patient's progression through stages of sepsis is critical for enabling risk stratification and adaptive, personalized treatment. However, commonly used sepsis diagnostic criteria fail to account for significant underlying heterogeneity, both between patients as well as over time in a single patient. We introduce a hidden Markov model of sepsis progression that explicitly accounts for patient heterogeneity. Benchmarked against two sepsis diagnostic criteria, the model provides a useful tool to uncover a patient's latent sepsis trajectory and to identify high-risk patients in whom more aggressive therapy may be indicated., Comment: Accepted to NIPS ML4H 2017

Many machine learning problems can be formulated as consensus optimization problems which can be solved efficiently via a cooperative multi-agent system. However, the agents in the system can be unreliable due to a variety of reasons: noise, faults and attacks. Providing erroneous updates leads the optimization process in a wrong direction, and degrades the performance of distributed machine learning algorithms. This paper considers the problem of decentralized learning using ADMM in the presence of unreliable agents. First, we rigorously analyze the effect of erroneous updates (in ADMM learning iterations) on the convergence behavior of multi-agent system. We show that the algorithm linearly converges to a neighborhood of the optimal solution under certain conditions and characterize the neighborhood size analytically. Next, we provide guidelines for network design to achieve a faster convergence. We also provide conditions on the erroneous updates for exact convergence to the optimal solution. Finally, to mitigate the influence of unreliable agents, we propose \textsf{ROAD}, a robust variant of ADMM, and show its resilience to unreliable agents with an exact convergence to the optimum., Comment: 23 pages

Recent literature has shown that the control of False Discovery Rate (FDR) for distributed detection in wireless sensor networks (WSNs) can provide substantial improvement in detection performance over conventional design methodologies. In this paper, we further investigate system design issues in FDR based distributed detection. We demonstrate that improved system design may be achieved by employing the Kolmogorov-Smirnov distance metric instead of the deflection coefficient, as originally proposed in Ray&VarshneyAES11. We also analyze the performance of FDR based distributed detection in the presence of Byzantines. Byzantines are malicious sensors which send falsified information to the Fusion Center (FC) to deteriorate system performance. We provide analytical and simulation results on the global detection probability as a function of the fraction of Byzantines in the network. It is observed that the detection performance degrades considerably when the fraction of Byzantines is large. Hence, we propose an adaptive algorithm at the FC which learns the Byzantines' behavior over time and changes the FDR parameter to overcome the loss in detection performance. Detailed simulation results are provided to demonstrate the robustness of the proposed adaptive algorithm to Byzantine attacks in WSNs., Comment: 28 pages, 9 figures, to appear in IEEE Transactions on Aerospace and Electronic Systems