Generative models have demonstrated remarkable abilities in generating high-fidelity visual content. In this work, we explore how generative models can further be used not only to synthesize visual content but also to understand the properties of a scene given a natural image. We formulate scene understanding as an inverse generative modeling problem, where we seek to find conditional parameters of a visual generative model to best fit a given natural image. To enable this procedure to infer scene structure from images substantially different than those seen during training, we further propose to build this visual generative model compositionally from smaller models over pieces of a scene. We illustrate how this procedure enables us to infer the set of objects in a scene, enabling robust generalization to new test scenes with an increased number of objects of new shapes. We further illustrate how this enables us to infer global scene factors, likewise enabling robust generalization to new scenes. Finally, we illustrate how this approach can be directly applied to existing pretrained text-to-image generative models for zero-shot multi-object perception. Code and visualizations are at https://energy-based-model.github.io/compositional-inference.

Industrial assembly lines are the heartbeat of modern manufacturing, where precision and efficiency are paramount. This paper introduces a novel hybrid Explainable artificial intelligence (XAI) approach to enhance monitoring and analysis in industrial assembly. By fusing the power of vision anomaly detection models with the clarity of the gradient tree boosting algorithm, this framework not only boosts defect detection accuracy but also provides transparent, actionable insights. This synergy transforms how operators and engineers interact with AI, fostering trust and enhancing operational excellence.

Even though virtual testing of Autonomous Vehicles (AVs) has been well recognized as essential for safety assessment, AV simulators are still undergoing active development. One particular challenge is the problem of including the Sensing and Perception (S&P) subsystem into the virtual simulation loop in an efficient and effective manner. In this article, we define Perception Error Models (PEM), a virtual simulation component that can enable the analysis of the impact of perception errors on AV safety, without the need to model the sensors themselves. We propose a generalized data-driven procedure towards parametric modeling and evaluate it using Apollo, an open-source driving software, and nuScenes, a public AV dataset. Additionally, we implement PEMs in SVL, an open-source vehicle simulator. Furthermore, we demonstrate the usefulness of PEM-based virtual tests, by evaluating camera, LiDAR, and camera-LiDAR setups. Our virtual tests highlight limitations in the current evaluation metrics, and the proposed approach can help study the impact of perception errors on AV safety.

This article proposes a generative adversarial network called explicit affine disentangled generative adversarial network (EAD-GAN), which explicitly disentangles affine transform in a self-supervised manner. We propose an affine transform regularizer to force the InfoGAN to have explicit properties of affine transform. To facilitate training an affine transform encoder, we decompose the affine matrix into two separate matrices and infer the explicit transform parameters by the least-squares method. Unlike the existing approaches, representations learned by the proposed EAD-GAN have clear physical meaning, where transforms, such as rotation, horizontal and vertical zooms, skews, and translations, are explicitly learned from training data. Thus, we set different values of each transform parameter individually to generate specifically affine transformed data by the learned network. We show that the proposed EAD-GAN successfully disentangles these attributes on the MNIST, CelebA, and dSprites datasets. EAD-GAN achieves higher disentanglement scores with a large margin compared to the state-of-the-art methods on the dSprites dataset. For example, on the dSprites dataset, EAD-GAN achieves the MIG and DCI score of 0.59 and 0.96 respectively, compared to 0.37 and 0.71, respectively, for the state-of-the-art methods.

Nowcasting is an observation-based method that uses the current state of the atmosphere to forecast future weather conditions over several hours. Recent studies have shown the promising potential of using deep learning models for precipitation nowcasting. In this paper, novel deep generative models are proposed for precipitation nowcasting. These models are equipped with extreme-value losses to more reliably predict extreme precipitation events. The proposed deep generative model contains a Vector Quantization Generative Adversarial Network and a Transformer ("VQGAN + Transformer"). For enhanced modeling and forecasting of extreme events, Extreme Value Loss (EVL) is incorporated in the autore-gressive Transformer. The numerical results show that the proposed model achieves comparable performance with the state-of-the-art conventional nowcasting method PySTEPS for predicting nominal values. By incorporating an EVL, the proposed model yields more accurate nowcasting of extreme precipitation.

In this paper, we aim to design an automatic camera pose estimation pipeline for clinical spaces such as catheterization laboratories. Our proposed pipeline exploits Scaled-YOLOv4 to detect fixed objects. We adopt the self-supervised key-point detector SuperPoint in combination with SuperGlue, a keypoint matching technique based on graph neural networks. Thus, we match key-points on input images with annotated reference points. Reference points are chosen on fixed objects in the scene, such as corners of door posts or windows. The point-correspondences between the image coordinates and the 3D coordinates are applied to the Perspective-n-Point algorithm to estimate the pose of each camera. Compared with other camera pose estimation methods, the proposed pipeline does not require the construction of 3D point-cloud model of the scene or placing a polyhedron object in the scene before each required calibration. Using videos from real procedures, we show that the pipeline can estimate the camera pose with high accuracy.

Slot attention has shown remarkable object-centric representation learning performance in computer vision tasks without requiring any supervision. Despite its object-centric binding ability brought by compositional modelling, as a deterministic module, slot attention lacks the ability to generate novel scenes. In this paper, we propose the Slot-VAE, a generative model that integrates slot attention with the hierarchical VAE framework for object-centric structured scene generation. For each image, the model simultaneously infers a global scene representation to capture high-level scene structure and object-centric slot representations to embed individual object components. During generation, slot representations are generated from the global scene representation to ensure coherent scene structures. Our extensive evaluation of the scene generation ability indicates that Slot-VAE outperforms slot representation-based generative baselines in terms of sample quality and scene structure accuracy.

To enhance the reliability and robustness of language identification (LID) and language diarization (LD) systems for heterogeneous populations and scenarios, there is a need for speech processing models to be trained on datasets that feature diverse language registers and speech patterns. We present the MERLIon CCS challenge, featuring a first-of-its-kind Zoom video call dataset of parent-child shared book reading, of over 30 hours with over 300 recordings, annotated by multilingual transcribers using a high-fidelity linguistic transcription protocol. The audio corpus features spontaneous and in-the-wild English-Mandarin code-switching, child-directed speech in non-standard accents with diverse language-mixing patterns recorded in a variety of home environments. This report describes the corpus, as well as LID and LD results for our baseline and several systems submitted to the MERLIon CCS challenge using the corpus.

Language development experts need tools that can automatically identify languages from fluent, conversational speech and provide reliable estimates of usage rates at the level of an individual recording. However, LID systems are typically evaluated on metrics such as equal error rate and balanced accuracy, applied at the level of an entire speech corpus. These overview metrics do not provide information about model performance at the level of individual speakers, recordings, or units of speech with different linguistic characteristics. Overview statistics may mask systematic errors in model performance for some subsets of the data, and consequently, have worse performance on data derived from some subsets of human speakers, creating a kind of algorithmic bias. Here, we investigate how well a number of LID systems perform on individual recordings and speech units with different linguistic properties in the MERLIon CCS Challenge featuring accented code-switched child-directed speech.

Neurologists typically identify epileptic seizures from electroencephalograms (EEGs) by visual inspection. This process is often time-consuming, especially for EEG recordings that last hours or days. To expedite the process, a reliable, automated, and patient-independent seizure detector is essential. However, developing a patient-independent seizure detector is challenging as seizures exhibit diverse characteristics across patients and recording devices. In this study, we propose a patient-independent seizure detector to automatically detect seizures in both scalp EEG and intracranial EEG (iEEG). First, we deploy a convolutional neural network with transformers and belief matching loss to detect seizures in single-channel EEG segments. Next, we extract regional features from the channel-level outputs to detect seizures in multi-channel EEG segments. At last, we apply post-processing filters to the segment-level outputs to determine seizures' start and end points in multi-channel EEGs. Finally, we introduce the minimum overlap evaluation scoring as an evaluation metric that accounts for minimum overlap between the detection and seizure, improving upon existing assessment metrics. We trained the seizure detector on the Temple University Hospital Seizure (TUH-SZ) dataset and evaluated it on five independent EEG datasets. We evaluate the systems with the following metrics: sensitivity (SEN), precision (PRE), and average and median false positive rate per hour (aFPR/h and mFPR/h). Across four adult scalp EEG and iEEG datasets, we obtained SEN of 0.617-1.00, PRE of 0.534-1.00, aFPR/h of 0.425-2.002, and mFPR/h of 0-1.003. The proposed seizure detector can detect seizures in adult EEGs and takes less than 15[Formula: see text]s for a 30[Formula: see text]min EEG. Hence, this system could aid clinicians in reliably identifying seizures expeditiously, allocating more time for devising proper treatment.

Estimating a sequence of dynamic undirected graphical models, in which adjacent graphs share similar structures, is of paramount importance in various social, financial, biological, and engineering systems, since the evolution of such networks can be utilized for example to spot trends, detect anomalies, predict vulnerability, and evaluate the impact of interventions. Existing methods for learning dynamic graphical models require the tuning parameters that control the graph sparsity and the temporal smoothness to be selected via brute-force grid search. Furthermore, these methods are computationally burdensome with time complexity$\mathcal {O}(NP^3)$for$P$variables and$N$time points. As a remedy, we propose a low-complexity tuning-free Bayesian approach, named BASS. Specifically, we impose temporally dependent spike and slab priors on the graphs such that they are sparse and varying smoothly across time. An efficient variational inference algorithm based on natural gradients is then derived to learn the graph structures from the data in an automatic manner. Owing to the pseudo-likelihood and the mean-field approximation, the time complexity of BASS is only$\mathcal {O}(NP^2)$. To cope with the local maxima problem of variational inference, we resort to simulated annealing and propose a method based on bootstrapping of the observations to generate the annealing noise. We provide numerical evidence that BASS outperforms existing methods on synthetic data in terms of structure estimation, while being more efficient especially when the dimension$P$becomes high. We further apply the approach to the stock return data of 78 banks from 2005 to 2013 and find that the number of edges in the financial network as a function of time contains three peaks, in coincidence with the 2008 global financial crisis and the two subsequent European debt crisis. On the other hand, by identifying the frequency-domain resemblance to the time-varying graphical models, we show that BASS can be extended to learning frequency-varying inverse spectral density matrices, and further yields graphical models for multivariate stationary time series. As an illustration, we analyze scalp EEG signals of patients at the early stages of Alzheimer’s disease (AD) and show that the brain networks extracted by BASS can better distinguish between the patients and the healthy controls.

We propose a manager-worker framework (the implementation of our model is publically available at:https://github.com/zcaicaros/manager-worker-mtsptwr) based on deep reinforcement learning to tackle a hard yet nontrivial variant of Travelling Salesman Problem (TSP), i.e., multiple-vehicle TSP with time window and rejections (mTSPTWR), where customers who cannot be served before the deadline are subject to rejections. Particularly, in the proposed framework, a manager agent learns to divide mTSPTWR into sub-routing tasks by assigning customers to each vehicle via a Graph Isomorphism Network (GIN) based policy network. A worker agent learns to solve sub-routing tasks by minimizing the cost in terms of both tour length and rejection rate for each vehicle, the maximum of which is then fed back to the manager agent to learn better assignments. Experimental results demonstrate that the proposed framework outperforms strong baselines in terms of higher solution quality and shorter computation time. More importantly, the trained agents also achieve competitive performance for solving unseen larger instances.

The worldwide development of Autonomous Vehicles (AVs) has also encouraged the use of software simulators for virtual testing of AVs. However, the effectiveness of the AV simulators is constrained by numerous challenges, such as their computational cost and lack of fidelity in specific areas. In this paper, we describe the modality of virtual testing and its benefits for AV development and validation. Moreover, we summarize and provide an overview of the state-of-the-art AV simulators, their limitations, and the current directions toward improvement.

State-of-the-art image denoisers exploit various types of deep neural networks via deterministic training. Alternatively, very recent works utilize deep reinforcement learning for restoring images with diverse or unknown corruptions. Though deep reinforcement learning can generate effective policy networks for operator selection or architecture search in image restoration, how it is connected to the classic deterministic training in solving inverse problems remains unclear. In this work, we propose a novel image denoising scheme via Residual Recovery using Reinforcement Learning, dubbed R3L. We show that R3L is equivalent to a deep recurrent neural network that is trained using a stochastic reward, in contrast to many popular denoisers using supervised learning with deterministic losses. To benchmark the effectiveness of reinforcement learning in R3L, we train a recurrent neural network with the same architecture for residual recovery using the deterministic loss, thus to analyze how the two different training strategies affect the denoising performance. With such a unified benchmarking system, we demonstrate that the proposed R3L has better generalizability and robustness in image denoising when the estimated noise level varies, comparing to its counterparts using deterministic training, as well as various state-of the-art image denoising algorithms.

We propose two end-to-end neural configurations for language diarization on bilingual code-switching speech. The first, a BLSTM-E2E architecture, includes a set of stacked bidirectional LSTMs to compute embeddings and incorporates the deep clustering loss to enforce grouping of languages belonging to the same class. The second, an XSA-E2E architecture, is based on an x-vector model followed by a self-attention encoder. The former encodes frame-level features into segmentlevel embeddings while the latter considers all those embeddings to generate a sequence of segment-level language labels. We evaluated the proposed methods on the dataset obtained from the shared task B in WSTCSMC 2020 and our handcrafted simulated data from the SEAME dataset. Experimental results show that our proposed XSA-E2E architecture achieved a relative improvement of 12.1% in equal error rate and a 7.4% relative improvement on accuracy compared with the baseline algorithm in the WSTCSMC 2020 dataset. Our proposed XSA-E2E architecture achieved an accuracy of 89.84% with a baseline of 85.60% on the simulated data derived from the SEAME dataset.

Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Pathological slowing in the electroencephalogram (EEG) is widely investigated for the diagnosis of neurological disorders. Currently, the gold standard for slowing detection is the visual inspection of the EEG by experts, which is time-consuming and subjective. To address those issues, we propose three automated approaches to detect slowing in EEG: Threshold-based Detection System (TDS), Shallow Learning-based Detection System (SLDS), and Deep Learning-based Detection System (DLDS). These systems are evaluated on channel-, segment-, and EEG-level. The three systems perform prediction via detecting slowing at individual channels, and those detections are arranged in histograms for detection of slowing at the segment- and EEG-level. We evaluate the systems through Leave-One-Subject-Out (LOSO) cross-validation (CV) and Leave-One-Institution-Out (LOIO) CV on four datasets from the US, Singapore, and India. The DLDS achieved the best overall results: LOIO CV mean balanced accuracy (BAC) of 71.9%, 75.5%, and 82.0% at channel-, segment- and EEG-level, and LOSO CV mean BAC of 73.6%, 77.2%, and 81.8% at channel-, segment-, and EEG-level. The channel- and segment-level performance is comparable to the intra-rater agreement (IRA) of an expert of 72.4% and 82%. The DLDS can process a 30 min EEG in 4 s and can be deployed to assist clinicians in interpreting EEGs.

Epilepsy diagnosis based on Interictal Epileptiform Discharges (IEDs) in scalp electroencephalograms (EEGs) is laborious and often subjective. Therefore, it is necessary to build an effective IED detector and an automatic method to classify IED-free versus IED EEGs. In this study, we evaluate features that may provide reliable IED detection and EEG classification. Specifically, we investigate the IED detector based on convolutional neural network (ConvNet) with different input features (temporal, spectral, and wavelet features). We explore different ConvNet architectures and types, including 1D (one-dimensional) ConvNet, 2D (two-dimensional) ConvNet, and noise injection at various layers. We evaluate the EEG classification performance on five independent datasets. The 1D ConvNet with preprocessed full-frequency EEG signal and frequency bands (delta, theta, alpha, beta) with Gaussian additive noise at the output layer achieved the best IED detection results with a false detection rate of 0.23/min at 90% sensitivity. The EEG classification system obtained a mean EEG classification Leave-One-Institution-Out (LOIO) cross-validation (CV) balanced accuracy (BAC) of 78.1% (area under the curve (AUC) of 0.839) and Leave-One-Subject-Out (LOSO) CV BAC of 79.5% (AUC of 0.856). Since the proposed classification system only takes a few seconds to analyze a 30-min routine EEG, it may help in reducing the human effort required for epilepsy diagnosis.

The diagnosis of epilepsy often relies on a reading of routine scalp electroencephalograms (EEGs). Since seizures are highly unlikely to be detected in a routine scalp EEG, the primary diagnosis depends heavily on the visual evaluation of Interictal Epileptiform Discharges (IEDs). This process is tedious, expert-centered, and delays the treatment plan. Consequently, the development of an automated, fast, and reliable epileptic EEG diagnostic system is essential. In this study, we propose a system to classify EEG as epileptic or normal based on multiple modalities extracted from the interictal EEG. The ensemble system consists of three components: a Convolutional Neural Network (CNN)-based IED detector, a Template Matching (TM)-based IED detector, and a spectral feature-based classifier. We evaluate the system on datasets from six centers from the USA, Singapore, and India. The system yields a mean Leave-One-Institution-Out (LOIO) cross-validation (CV) area under curve (AUC) of 0.826 (balanced accuracy (BAC) of 76.1%) and Leave-One-Subject-Out (LOSO) CV AUC of 0.812 (BAC of 74.8%). The LOIO results are found to be similar to the interrater agreement (IRA) reported in the literature for epileptic EEG classification. Moreover, as the proposed system can process routine EEGs in a few seconds, it may aid the clinicians in diagnosing epilepsy efficiently.

Smooth handling of pedestrian interactions is a key requirement for Autonomous Vehicles (AV) and Advanced Driver Assistance Systems (ADAS). Such systems call for early and accurate prediction of a pedestrian’s crossing/not-crossing behaviour in front of the vehicle. Existing approaches to pedestrian behaviour prediction make use of pedestrian motion, his/her location in a scene and static context variables such as traffic lights, zebra crossings etc. We stress on the necessity of early prediction for smooth operation of such systems. We introduce the influence of vehicle interactions on pedestrian intention for this purpose. In this paper, we show a discernible advance in prediction time aided by the inclusion of such vehicle interaction context. We apply our methods to two different datasets, one in-house collected - NTU dataset and another public real-life benchmark - JAAD dataset. We also propose a generalization of the Latent-Dynamic Conditional Random Fields (LDCRF), called Factored LDCRF (FLDCRF), for improved sequence prediction performance. FLDCRF outperforms Long Short-Term Memory (LSTM) networks across the datasets over identical time-series features. While the existing best system predicts pedestrian stopping behaviour with 70% accuracy 0.38 seconds before the actual events, our system achieves such accuracy at least 0.9 seconds on an average before the actual events across datasets.

Accelerating the performance of optimization algorithms is crucial for many day-to-day applications. Mobility-on-demand is one such application that is transforming urban mobility by offering reliable and convenient on-demand door-to-door transportation at any time. Dial-a-ride problem (DARP) is an underlying optimization problem in the operational planning of mobility-on-demand systems. The primary objective of DARP is to design routes and schedules to serve passenger transportation requests with high-level user comfort. DARP often arises in dynamic real-world scenarios, where rapid route planning is essential. The traditional CPU-based algorithms are generally too slow to be useful in practice. Since customers expect quick response for their mobility requests, there has been a growing interest in fast solution methods. Therefore, in this paper, we introduce a GPU-based solution methodology for the dial-a-ride problem to produce good solutions in a short time. Specifically, we develop a GPU framework to accelerate time-critical neighborhood exploration of local search operations under the guidance of metaheuristics such as tabu search and variable neighborhood search. Besides, we propose device-oriented optimization strategies to enhance the utilization of a current-generation GPU architecture (Tesla P100). We report speedup achieved by our GPU approach when compared to its classical CPU counterpart, and the effect of each device optimization strategy on computational speedup. Results are based on standard test instances from the literature. Ultimately, the proposed GPU methodology generates better solutions in a short time when compared to the existing sequential approaches.

The log-determinant optimization problem with general matrix constraints arises in many applications. The log-determinant term hampers the scalability of existing methods. This paper proposes a highly efficient stochastic method that has time complexity O(N2), whereas existing methods have complexity O(N3). In order to achieve the quadratic complexity, the proposed algorithm leverages an efficient stochastic gradient of the augmented Lagrangian form and relies on subgradient descent method. Convergence of this method is analyzed both theoretically and empirically. The resulting primal-dual stochastic subgradient method yields the same accuracy as existing methods yet only requires O(N2) operations.

Sensing and Perception (S&P) is a crucial component of an autonomous system (such as a robot), especially when deployed in highly dynamic environments where it is required to react to unexpected situations. This is particularly true in case of Autonomous Vehicles (AVs) driving on public roads. However, the current evaluation metrics for perception algorithms are typically designed to measure their accuracy per se and do not account for their impact on the decision making subsystem(s). This limitation does not help developers and third party evaluators to answer a critical question: is the performance of a perception subsystem sufficient for the decision making subsystem to make robust, safe decisions? In this paper, we propose a simulation-based methodology towards answering this question. At the same time, we show how to analyze the impact of different kinds of sensing and perception errors on the behavior of the autonomous system.

Recently deep reinforcement learning has shown success in solving NP-hard combinatorial optimization problems such as traveling salesman problems, vehicle routing problems, job-shop scheduling problems, as well as their variants. However, most of the problems being solved are still relatively simple compared to the real-world scenarios. For instance, feasibility constraints are rarely considered in the current frameworks. This paper investigates the possibility of applying deep reinforcement learning to tackle combinatorial optimization problems with feasibility constraints. We propose a framework to solve such problems by combining deep reinforcement learning with a greedy heuristic. We demonstrate this approach for the traveling salesman problem with time windows and rejection (TSPTWR). The results show that our approach outperforms a commonly employed tabu search heuristic, both in terms of the solution quality and the inference computation time. More specifically, the inference process is 100 to 1000 times faster than tabu search for different size TSPTWR. The proposed approach can be considered as a framework enhancing reinforcement learning with heuristics for solving more complex problems.

Visual evaluation of electroencephalogram (EEG) for Interictal Epileptiform Discharges (IEDs) as distinctive biomarkers of epilepsy has various limitations, including time-consuming reviews, steep learning curves, interobserver variability, and the need for specialized experts. The development of an automated IED detector is necessary to provide a faster and reliable diagnosis of epilepsy. In this paper, we propose an automated IED detector based on Convolutional Neural Networks (CNNs). We have evaluated the proposed IED detector on a sizable database of 554 scalp EEG recordings (84 epileptic patients and 461 nonepileptic subjects) recorded at Massachusetts General Hospital (MGH), Boston. The proposed CNN IED detector has achieved superior performance in comparison with conventional methods with a mean cross-validation area under the precision-recall curve (AUPRC) of 0.838[Formula: see text]±[Formula: see text]0.040 and false detection rate of 0.2[Formula: see text]±[Formula: see text]0.11 per minute for a sensitivity of 80%. We demonstrated the proposed system to be noninferior to 30 neurologists on a dataset from the Medical University of South Carolina (MUSC). Further, we clinically validated the system at National University Hospital (NUH), Singapore, with an agreement accuracy of 81.41% with a clinical expert. Moreover, the proposed system can be applied to EEG recordings with any arbitrary number of channels.

Structure learning of Gaussian graphical models typically involves careful tuning of penalty parameters, which balance the tradeoff between data fidelity and graph sparsity. Unfortunately, this tuning is often a “black art” requiring expert experience or brute-force search. It is therefore tempting to develop tuning-free algorithms that can determine the sparsity of the graph adaptively from the observed data in an automatic fashion. In this paper, we propose a novel approach, named BISN (Bayesian inference of Sparse Networks), for automatic Gaussian graphical model selection. Specifically, we regard the off-diagonal entries in the precision matrix as random variables and impose sparse-promoting horseshoe priors on them, resulting in automatic sparsity determination. With the help of stochastic gradients, an efficient variational Bayes algorithm is derived to learn the model. We further propose a decaying recursive stochastic gradient (DRSG) method to reduce the variance of the stochastic gradients and to accelerate the convergence. Our theoretical analysis shows that the time complexity of BISN scales only quadratically with the dimension, whereas the theoretical time complexity of the state-of-the-art methods for automatic graphical model selection is typically a third-order function of the dimension. Furthermore, numerical results show that BISN can achieve comparable or better performance than the state-of-the-art methods in terms of structure recovery, and yet its computational time is several orders of magnitude shorter, especially for large dimensions.

Travel time estimation is an important aspect of intelligent transportation systems (ITS). In urban environments, travel times can exhibit much variability due to various stochastic factors. For this reason, we focus on estimating travel time distributions, in contrast to the more commonly studied estimation of mean expected travel times. We present algorithms to infer travel time distributions from Floating Car Data; specifically, from sparse GPS measurements. The framework combines Gaussian copulas and network inference to estimate marginal and joint distributions of travel times. We perform an extensive set of numerical experiments on one month of GPS trajectories. We benchmark the proposed models in terms of Kullback-Leibler (KL) divergence and Hellinger distance for the 50 most common trajectories. Combining Gaussian Copulas and Bayesian Inference of Sparse Networks method achieves 4.9% reduction in KL divergence and 2% reduction in Hellinger distance compared to baseline methods.

We consider the structure learning problem of Gaussian graphical models when the underlying graph is semi-sparse. More specifically, we assume that the number of edges in the graph grows quadratically with the dimension P. Similar to the case of sparse graphs, the problem is formulated as maximizing the data log-likelihood with an ℓ1norm penalty on the precision matrix (the inverse covariance matrix) that promotes sparsity. We notice that the time complexity of all existing methods is at least O(P3) under the scenario of semi-sparse graphs, thus severely hindering their applications to high-dimensional data. By contrast, the time complexity of the proposed method is only O(P2) with the help of stochastic gradients. We prove the convergence of the proposed algorithm. Numerical results show that the computational time of the proposed method is shorter than that of the state-of-the-art methods when the graph is semi-sparse.

The motivation for this paper is derived from the fact that there has been increasing interest among researchers and practitioners in developing technologies that capture, model and analyze learning and teaching experiences that take place beyond computer-based learning environments. In this paper, we review case studies of tools and technologies developed to collect and analyze data in educational settings, quantify learning and teaching processes and support assessment of learning and teaching in an automated fashion. We focus on pipelines that leverage information and data harnessed from physical spaces and/or integrates collected data across physical and digital spaces. Our review reveals a promising field of physical classroom analysis. We describe some trends and suggest potential future directions. Specifically, more research should be geared towards a) deployable and sustainable data collection set-ups in physical learning environments, b) teacher assessment, c) developing feedback and visualization systems and d) promoting inclusivity and generalizability of models across populations.

Human-robot interaction in corporate workplaces is a research area which remains unexplored. In this paper, we present the results and analysis of a social experiment we conducted by introducing a humanoid robot (Nadine) into a collaborative social workplace. The humanoid's primary task was to function as a receptionist and provide general assistance to the customers. Moreover, the employees who interacted with Nadine were given over a month to get used to her capabilities, after which, the feedback was collected from the staff on the grounds of influence on productivity, affect experienced during interaction and their views on social robots assisting with regular tasks. Our results show that the usage of social robots for assisting with normal day-to-day tasks is taken quite positively by the co-workers and that in the near future, more capable humanoid social robots can be used in workplaces for assisting with menial tasks. Finally, we posit that surveys such as ours could result in constructive opinions based on technological awareness, rather than opinions from media-driven fears about the threats of technology.

Point set registration presents unique significance in Lidar-based intelligent vehicle localization and mapping. It involves registering point sets of the same scene observed from different positions by determining their relative spatial transformation. However, due to the noise and outliers in the point sets and initial misalignment, existing methods suffer from the issues of low accuracy or large computational cost. In this paper, we propose a novel Bayesian state space model to describe the sequential point registration problem. Specifically, we specify the transformations to be the latent states and further assume that they vary smoothly across time. The point clouds are then represented as Gaussian mixture models that change accordingly with the transformation. We then develop a stochastic variational Bayesian inference algorithm to learning the distributions of the transformation, which automatically strike a balance between mapping every two consecutive point clouds and the temporal smoothness of the transformation. Experimental results based simulated data show that the proposed variational Bayesian point set registration (VB-PSR) algorithm achieves higher accuracy with comparable or less time and resources, in comparison with the state- of-the-art methods.

The simultaneous localization and mapping (SLAM) problem involves using the measurements of sensors to construct an environmental map, while simultaneously recovering the vehicle trajectory within this map. There are broadly two strategies for SLAM: on-line and off-line. In this paper, we focus on the off-line SLAM (a.k.a. full SLAM) problem and propose a variational Bayes inference algorithm to address it. Specifically, the intractable posterior distribution of the vehicle poses given the measurements is approximated by a tractable variational distribution, resulting in estimates of the vehicle poses as well as their uncertainties. In contrast with the existing off- line methods, the inverse variances of the additive noise are updated along with the posterior distribution instead of being fixed, thus, the proposed method is robust to unknown noises. Furthermore, the computational complexity of the proposed method is only linear in the number of frames and the computational bottleneck of the algorithm can be easily parallelized to achieve further acceleration. Numerical results show that the proposed method is insensitive to the selection of the noise parameters. More importantly, it is superior in efficiency to the state-of-the-art method, especially for large- scale SLAM problems.

In recent years, there is a sharp increase in the PMD (Personal Mobility Device) usage for first and last mile commute in developed nations such as Singapore. However, there is a lack of dedicated infrastructure for PMDs which forces them to share the path with pedestrians. This increases the safety risk for PMD riders as well as the pedestrians. This paper explains a practical experiment carried out to understand the reaction of the pedestrians to the PMDs. The effect of the factors such as speed of the PMD, passing distance, the gender of the pedestrians and the PMD model itself are considered in the pedestrian perception results. The results indicate that pedestrians are cautious to a newer PMD model and are more comfortable when they see a familiar PMD model. In addition to that, the effect of pedestrian gender on perception results is discussed. A sample prediction application is described which can alert the PMD riders if they are causing discomfort to the pedestrians in their path. The findings of our study validate the recent safety guidelines and speed limit regulations passed by Singapore's government for PMD users.

This paper addresses the issue of video-based action recognition by exploiting an advanced multistream convolutional neural network (CNN) to fully use semantics-derived multiple modalities in both spatial (appearance) and temporal (motion) domains, since the performance of the CNN-based action recognition methods heavily relates to two factors: semantic visual cues and the network architecture. Our work consists of two major parts. First, to extract useful human-related semantics accurately, we propose a novel spatiotemporal saliency-based video object segmentation (STS) model. By fusing different distinctive saliency maps, which are computed according to object signatures of complementary object detection approaches, a refined STS maps can be obtained. In this way, various challenges in the realistic video can be handled jointly. Based on the estimated saliency maps, an energy function is constructed to segment two semantic cues: the actor and one distinctive acting part of the actor. Second, we modify the architecture of the two-stream network (TS-Net) to design a multistream network that consists of three TS-Nets with respect to the extracted semantics, which is able to use deeper abstract visual features of multimodalities in multi-scale spatiotemporally. Importantly, the performance of action recognition is significantly boosted when integrating the captured human-related semantics into our framework. Experiments on four public benchmarks-JHMDB, HMDB51, UCF-Sports, and UCF101-demonstrate that the proposed method outperforms the state-of-the-art algorithms.

Despite outstanding performance in image recognition, convolutional neural networks (CNNs) do not yet achieve the same impressive results on action recognition in videos. This is partially due to the inability of CNN for modeling long-range temporal structures especially those involving individual action stages that are critical to human action recognition. In this paper, we propose a novel action-stage (ActionS) emphasized spatiotemporal Vector of Locally Aggregated Descriptors (ActionS-STVLAD) method to aggregate informative deep features across the entire video according to adaptive video feature segmentation and adaptive segment feature sampling (AVFS-ASFS). In our ActionSST- VLAD encoding approach, by using AVFS-ASFS, the key frame features are chosen and the corresponding deep features are automatically split into segments with the features in each segment belonging to a temporally coherent ActionS. Then, based on the extracted key frame feature in each segment, a flow-guided warping technique is introduced to detect and discard redundant feature maps, while the informative ones are aggregated by using our exploited similarity weight. Furthermore, we exploit an RGBF modality to capture motion salient regions in the RGB images corresponding to action activity. Extensive experiments are conducted on four public benchmarks - HMDB51, UCF101, Kinetics and ActivityNet for evaluation. Results show that our method is able to effectively pool useful deep features spatiotemporally, leading to state-of-the-art performance for videobased action recognition.

Non-recurring incidents such as accidents and vehicle breakdowns are the leading causes of severe traffic congestions in large cities. Consequently, anticipating the duration of such events in advance can be highly useful in mitigating the resultant congestion. However, availability of partial information or ever-changing ground conditions makes the task of forecasting the duration particularly challenging. In this paper, we propose an adaptive ensemble model that can provide reasonable forecasts even when a limited amount of information is available and further improves the prediction accuracy as more information becomes available during the course of the incidents. Furthermore, we consider the scenarios where the historical incident reports may not always contain accurate information about the duration of the incidents. To mitigate this issue, we first quantify the effective duration of the incidents by looking for the change points in traffic state and then utilize this information to predict the duration of the incidents. We compare the prediction performance of different traditional regression methods, and the experimental results show that the Treebagger outperforms other methods. For the incidents with duration in the range of 36 - 200 min, the mean absolute percentage error (MAPE) in predicting the duration is in the range of 25% - 55%. Moreover, for longer duration incidents (greater than 65 min), prediction improves significantly with time. For example, the MAPE value varies over time from 76% to 50% for incidents having a duration greater than 200 min. Finally, the overall MAPE value averaged over all incidents improves by 50% with elapsed time for prediction of reported as well as effective duration.

Insider threats pose a challenge to all companies and organizations. Identification of culprit after an attack is often too late and result in detrimental consequences for the organization. Majority of past research on insider threat has focused on post-hoc personality analysis of known insider threats to identify personality vulnerabilities. It has been proposed that certain personality vulnerabilities place individuals to be at risk to perpetuating insider threats should the environment and opportunity arise. To that end, this study utilizes a game-based approach to simulate a scenario of intellectual property theft and investigate behavioral and personality differences of individuals who exhibit insider-threat related behavior. Features were extracted from games, text collected through implicit and explicit measures, simultaneous facial expression recordings, and personality variables (HEXACO, Dark Triad and Entitlement Attitudes) calculated from questionnaire. We applied ensemble machine learning algorithms and show that they produce an acceptable balance of precision and recall. Our results showcase the possibility of harnessing personality variables, facial expressions and linguistic features in the modeling and prediction of insider-threat.

Speech disorders are among the salient characteristics of negative symptoms of schizophrenia. Such impairments are often exhibited through disorganized speech, inappropriate affective prosody, and poverty of speech. The current method of detecting such symptoms requires the expertise of a trained clinician, which may be prohibitive due to cost, stigma or high patient-to-clinician ratio. An objective method to extract non-verbal and verbal speech-related cues can help to automate and simplify the assessment method of severity of speech-related symptoms of schizophrenia. In this paper, a novel automated method is presented which uses speech content from schizophrenic patients to predict the clinician-assigned subjective ratings of their negative symptoms. Specifically, the interviews of 50 schizophrenia patients were recorded and features related to acoustics, linguistics and non-verbal conversation were extracted. The subjective ratings can be accurately predicted from the objective features with an accuracy of 64-82% using machine learning algorithms with leave-one-out cross-validation. Our findings support the utility of automated speech analysis to aid clinician diagnosis, monitoring and understanding of schizophrenia.

Driving Heavy Duty Vehicles (HDVs) as a platoon has potential to significantly reduce the fuel consumption, human labor and increase the safety. A suitable controller which can maintain the vehicle movement in a defined topology is essential for HDV platooning. This paper proposes a controller based on the combination of Constant Distance (CD) and Headway Time (HT) topologies using Model Predictive Control (MPC) for a longitudinal HDVs platoon. In addition to this, a MPC controller is compared with conventional PID (Proportional-Integral-Derivative) controller. The controller aims to maintain intervehicular distance and headway time between the vehicles for two cases, namely unconstrained and constrained optimization problem. The predictive control algorithm uses a kinematic model of vehicle platooning. A systematic handling of constraints yields significant improvements in the performance of the proposed MPC strategy over conventional PID controller. A road network of a U.S. freeway I5 has been built in VISSIM for the simulations. The results and discussions are at the end of the paper.

The presence of Epileptiform Transients (ET) in the electroencephalogram (EEG) is a key finding in the medical workup of a patient with suspected epilepsy. Automated ET detection can increase the uniformity and speed of ET detection. Current ET detection methods suffer from insufficient precision and high false positive rates. Since ETs occur infrequently in the EEG of most patients, the majority of recordings comprise background EEG waveforms. In this work we establish a method to exclude as much background data as possible from EEG recordings by applying a classifier cascade. The remaining data can then be classified using other ET detection methods. We compare a single Support Vector Machine (SVM) to a cascade of SVMs for detecting ETs. Our results show that the precision and false positive rate improve significantly by incorporating a classifier cascade before ET detection. Our method can help improve the precision and false positive rate of an ET detection system. At a fixed sensitivity, we were able to improve precision by 6.78%; and at a fixed false positive rate, the sensitivity improved by 2.83%.

Negative symptoms of schizophrenia are often associated with the blunting of emotional affect which creates a serious impediment in the daily functioning of the patients. Affective prosody is almost always adversely impacted in such cases, and is known to exhibit itself through the low-level acoustic signals of prosody. To automate and simplify the process of assessment of severity of emotion related symptoms of schizophrenia, we utilized these low-level acoustic signals to predict the expert subjective ratings assigned by a trained psychologist during an interview with the patient. Specifically, we extract acoustic features related to emotion using the openSMILE toolkit from the audio recordings of the interviews. We analysed the interviews of 78 paid participants (52 patients and 26 healthy controls) in this study. The subjective ratings could be accurately predicted from the objective openSMILE acoustic signals with an accuracy of 61-85% using machine-learning algorithms with leave-one-out cross-validation technique. Furthermore, these objective measures can be reliably utilized to distinguish between the patient and healthy groups, as supervised learning methods can classify the two groups with 79-86% accuracy.

Semantic segmentation is an important task for autonomous vehicle navigation in off road environments. However, several natural factors make this problem uniquely challenging. For example, road segmentation is often difficult under heavy shadow or steel terrain, and dangerous muddy water puddles may have the similar visual appearance to dirt road surfaces (and thus are hard to identify). To tacule these challenges, we present a semantic segmentation system based on a two-stage hierarchical fusion pipeline. The first stage improves the road segmentation by effectively fusing information from camera and 3D Lidar point cloud. The second stage is dedicated to detecting water puddles, based on the results from the first stage. Due to the parallelized architecture, our system can be deployed for real-time applications. We achieved an F1 score of around 93% for road segmentation and 80% for water puddle segmentation at more than 10 Hz.

In underground low-pressure gas pipelines, a leak may result in a complication more dangerous and difficult to detect, known as the water ingress problem. Groundwater enters the gas pipeline through a crack, and eventually blocks the gas flow. Vibration sensors are used to detect the water ingress problem. The sensor output indicates a marked increase in the occurrence of spikes from the vibration sensor once the water ingress starts. Since the signal is not much greater in magnitude than the baseline signal it is difficult to definitively detect. The estimation error from a Kalman filter is used to detect the occurrence of water ingress and with multiple sensors even the location of the water ingress can be detected.

Many approaches for tracking objects in lidar data have been proposed in recent years. However, most practical real time systems assume that clean segmentation of lidar points into individual objects can be achieved. Unfortunately, efficient lidar segmentation approaches are prone to under-segmentation when objects are very close to each other; one solution is to introduce additional segmentation steps into the tracking process. In this paper we propose a new method to address this task with distance dependent Chinese Restaurant Processes (dd-CRP) equipped with a shape prior defining possible object shapes. By adding constraints to the segmentation model, we are able to further improve stability of segmentation and tracking. Experiments on real datasets show the advantage of this approach over a baseline object tracking pipeline.

Negative symptoms of schizophrenia significantly affect the daily functioning of patients, especially movement and expressive gestures. The diagnosis of such symptoms is often difficult and require the expertise of a trained clinician. Apart from these subjective methods, there is little research on developing objective methods to quantify the symptoms. Therefore, we explore body movement signals as objective measures of negative symptoms. Specifically, we extract the signals from video recordings of patients being interviewed. We analysed the interviews of 69 paid participants (46 patients and 23 healthy controls) in this study. Correlation between movement signals (linear and angular speeds of upper limbs and head, acceleration and gesture angles) and subjective ratings (assigned during same interview) from the NSA-16 scale were calculated. As hypothesized, the movement signals correlated strongly with the movement impairment aspect of the NSA-16 questionnaire. Also, not quite surprisingly, strong correlations were obtained between the movement signals and speech items of NSA-16, indicating lack of associated gestures in patients during speech. These subjective ratings could also be reasonably predicted from the objective signals with an accuracy of 61-78% using machine-learning algorithms with leave-one-out cross-validation technique. Furthermore, these objective measures can be reliably utilized to distinguish between the patient and healthy groups, as supervised learning methods can classify the two groups with 74-87% accuracy.

Brain-computer interface (BCI) is an emerging area of research that aims to improve the quality of human-computer applications. It has enormous scope in biomedical applications, neural rehabilitation, biometric authentication, educational programmes, and entertainment applications. A BCI system has four major components: signal acquisition, signal preprocessing, feature extraction, and classification. In this study, we provide a comparison of various traditional classification algorithms to the newer methods of deep learning. We explore two different types of deep learning methods, namely, convolutional neural networks (CNN) and recurrent neural networks (RNN) with long short-term memory (LSTM) architecture. We test the classification accuracies on a recent 5-dass steady-state visual evoked potential (SSVEP) dataset. The results prove the superiority of deep learning methods in comparison with the traditional classification algorithms. Amongst the traditional classifiers, support vector machine (SVM) with Gaussian kernel employing sequential forward selection (SFS) of features provided a better classification accuracy of 66.09%, while CNN provided the highest classification accuracy of 69.03%.

This paper is concerned with a system for detecting and tracking multiple 3D bounding boxes based on information from multiple sensors. Our framework is built around an inference engine similar to the probability hypothesis density (PHD) filter, where the state space consists of stochastic bounding boxes with constant velocity dynamics. We outline measurement equations for two modalities (vision and radar). The result is a flexible inference system suitable for use on autonomous vehicles.

Automated annotation of electroencephalograms (EEG) of epileptic patients is important in diagnosis and management of epilepsy. Epilepsy is often associated with the presence of epileptiform transients (ET) in the EEG. To develop an efficient ET detector, a vast amount of data is required to train and evaluate the performance of the detector. Interictal EEG data contains mostly background waveforms, since ETs only occur occasionally in most patients. In order to detect ETs in an automated fashion, it is meaningful to first try to eliminate most background waveforms by means of simple, fast classifiers. The remaining waveforms can in a following step be processed by more sophisticated and computationally demanding classification algorithms, such as deep learning systems. In this study, we design a cascade of simple thresholding steps to reject most background waveforms in interictal EEG, while maintaining most ETs. Several simple and quick-to-compute EEG features are chosen. By thresholding these features in consecutive steps, background waveforms are rejected sequentially. In our numerical experiments, a cascade of 10 steps is able to reject 98.65% of all background segments in the dataset, while preserving 90.6% of the ETs.

The EEG of epileptic patients often contains sharp waveforms called "spikes", occurring between seizures. Detecting such spikes is crucial for diagnosing epilepsy. In this paper, we develop a convolutional neural network (CNN) for detecting spikes in EEG of epileptic patients in an automated fashion. The CNN has a convolutional architecture with filters of various sizes applied to the input layer, leaky ReLUs as activation functions, and a sigmoid output layer. Balanced mini-batches were applied to handle the imbalance in the data set. Leave-one-patient-out cross-validation was carried out to test the CNN and benchmark models on EEG data of five epilepsy patients. We achieved 0.947 AUC for the CNN, while the best performing benchmark model, Support Vector Machines with Gaussian kernel, achieved an AUC of 0.912.

Negative symptoms in schizophrenia are associated with significant burden and functional impairment, especially speech production. In clinical practice today, there are no robust treatments for negative symptoms and one obstacle surrounding its research is the lack of an objective measure. To this end, we explore non-verbal speech cues as objective measures. Specifically, we extract these cues while schizophrenic patients are interviewed by psychologists. We have analyzed interviews of 15 patients who were enrolled in an observational study on the effectiveness of Cognitive Remediation Therapy (CRT). The subject (undergoing CRT) and control group (not undergoing CRT) contains 8 and 7 individuals respectively. The patients were recorded during three sessions while being evaluated for negative symptoms over a 12-week follow-up period. In order to validate the non-verbal speech cues, we computed their correlation with the Negative Symptom Assessment (NSA-16). Our results suggest a strong correlation between certain measures of the two rating sets. Supervised prediction of the subjective ratings from the non-verbal speech features with leave-one-person-out cross-validation has reasonable accuracy of 53-80%. Furthermore, the non-verbal cues can be used to reliably distinguish between the subjects and controls, as supervised learning methods can classify the two groups with 80-93% accuracy.

Epilepsy is often associated with the presence of spikes in electroencephalograms (EEGs). The spike waveforms vary vastly among epilepsy patients, and also for the same patient across time. In order to develop semi-automated and automated methods for detecting spikes, it is crucial to obtain a better understanding of the various spike shapes. In this paper, we develop several approaches to extract exemplars of spikes. We generate spike exemplars by applying clustering algorithms to a database of spikes from 12 patients. As similarity measures for clustering, we consider the Euclidean distance and Dynamic Time Warping (DTW). We assess two clustering algorithms, namely, K-means clustering and affinity propagation. The clustering methods are compared based on the mean squared error, and the similarity measures are assessed based on the number of generated spike clusters. Affinity propagation with DTW is shown to be the best combination for clustering epileptic spikes, since it generates fewer spike templates and does not require to pre-specify the number of spike templates.

A probabilistic method is proposed to detect and localize leaks in low-pressure gas distribution networks. These leakage events are estimated using flow and pressure information obtained from the steady state analysis of gas network. The approach provides an estimation of the leaked pipe section and the amount of gas outflow from the pipe section containing leaks. The reliability of the methodology is shown by analyzing the network in the presence of network modeling errors. These errors account for the variation in demand value of gas at the outlet nodes and the change in pipe roughness due to age. Even in the presence of large noise in the network, this methodology provides an accuracy of more than 80% in the localization of leak. The study aims to develop a real-time online monitoring system for low pressure gas distribution networks. Moreover, this technique is cost effective and can be easily integrated with the existing monitoring system.

Faults in a gas pipeline network is one of the major impairments towards the safe gas distribution among consumers. So, it is significant to detect and locate the faults in a pipeline network. In this work, a novel sensor based fault localization scheme, FLoGPN is proposed for detection and localization of faults in a gas distribution network. Here leak in gas pipeline network is considered as fault. This scheme uses Josang's Beta Reputation model to combine information from the sensors to select the nearest sensor to the fault. The aim is to identify the nearest sensor to the fault that indicates the faulty pipe segment in the whole network. The method is tested in the presence of noise to check its reliability and the simulation results show that the scheme can localize fault in a gas pipeline network accurately with the SNR value 20 dB or more.

Leak detection is vital in oil and gas pipelines, as it can cause financial loss and impact the environment, prove fatal to human life and also affect the effective functioning of domestic household. Acoustic emission test is a non-intrusive technique for leak detection. Leakage in pipes creates vibrations which are transmitted along the pipe walls. These waves can be detected by using acoustic sensors or accelerometers installed on the pipe wall for the purpose of analyses. In this work, a simulation study is carried where the leak location is defined by an Incident Pressure Field and this source propagates waves in the frequency range of about 20-200Hz. The results are verified by experimental work by introducing synthetic leak and disturbance signals. The Fast Fourier Transform (FFT) was used as the mathematical tool to evaluate the dominant peak in the frequency spectrum during an event of leak or disturbance. The experimental results obtained are in accordance with the simulated results. The accuracy of determining an event occurrence can be improved with error less than 5%.

Gas pipelines are the most vital and commonly used method of transportation of fuel, water, gas. Any form of leak in the pipeline can be catastrophic to mankind, incurring huge financial losses. In this paper, an experimental study has been carried out with respect to a real-time monitoring on the pressure and the flow variations. A small section of the pipeline with valves is considered to generate a synthetic leak in a pipeline. The regulation of these valves can be used to monitor the effects of leak on pressure and flow parameters of the town gas. Town gas is widely used in household and industrial applications in Singapore. A systematic test has been performed to analyze the irregularities due to leak. The performance and the magnitude of the signals will help in defining the effects due to leak location and valve operation.

Dynamic compressed sensing (DCS) has recently gained popularity as a successful approach to recovering dynamic sparse signals. In this paper, we attack the problem from a Bayesian perspective. The proposed model imposes sparse constraints on both the unknown sparse signal and its temporal innovation via t priors. Due to the conjugacy between the priors and likelihoods, we are able to propose a computationally efficient mean-field variational Bayes algorithm to learn the model without parameter tuning. We consider both the online and offline scenarios, and demonstrate via numerical experiments that the proposed methods are superior to alternatives in terms of both reconstruction accuracy and computational time.

We consider the problem of jointly describing extreme events at a multitude of locations, which presents paramount importance in catastrophes forecast and risk management. Specifically, a novel Ensemble-of-Latent-Trees of Pairwise Copula (ELTPC) model is proposed. In this model, the spatial dependence is captured by latent trees expressed by pairwise copulas. To compensate the limited expressiveness of every single latent tree, an mixture of latent trees is employed. By harnessing the variational inference and stochastic gradient techniques, we further develop a triply stochastic variational inference (TSVI) algorithm for learning and inference. The corresponding computational complexity is only linear in the number of variables. Numerical results from both the synthetic and real data show that the ELTPC model provides a reliable description of the spatial extremes in a flexible but parsimonious manner.

Social cognition is the study of how people interact with each other in a social situation. An effective interaction would require higher degree of cognitive involvement between the participants and consequently, an enhanced synchrony between their neural mechanisms. In this study, twelve pairs of subjects interacted with each other via a cognitively engaging experimental paradigm in which they either competed or cooperated with each other for performing a task. While they were performing the task we incorporated electroencephalographic (EEG) hyperscanning techniques by simultaneously recording the EEG activities of the interacting subjects. We quantified these interactions by computing the inter-brain synchrony (IBS) and studied the changes in IBS under different experimental conditions. We found that the inter-brain synchrony between the subjects was significantly higher when they cooperated with each other as compared to the competitive scenario. Furthermore, we found that IBS was significantly enhanced when the subjects were physically separated i.e. they cooperated via an intranet network. In this work, we have demonstrated how EEG hyperscanning technique can be employed to study inter-brain synchronization under different conditions.

Intelligent transportation systems (ITSs) gather information about traffic conditions by collecting data from a wide range of on-ground sensors. The collected data usually suffer from irregular spatial and temporal resolution. Consequently, missing data is a common problem faced by ITSs. In this paper, we consider the problem of missing data in large and diverse road networks. We propose various matrix and tensor based methods to estimate these missing values by extracting common traffic patterns in large road networks. To obtain these traffic patterns in the presence of missing data, we apply fixed-point continuation with approximate singular value decomposition, canonical polyadic decomposition, least squares, and variational Bayesian principal component analysis. For analysis, we consider different road networks, each of which is composed of around 1500 road segments. We evaluate the performance of these methods in terms of estimation accuracy, variance of the data set, and the bias imparted by these methods.

The role of smartphones in traffic applications is typically limited to front-end interface. Although smartphones have significant computational resources, which are most likely to increase further in the near future, most of the computations are still performed on servers. In this paper, we study the computational performance of centralized, decentralized, and hybrid architectures for intelligent transportation system applications. We test these architectures on various Android devices. For implementation, we consider Android Software Development Kit (SDK) and Android Native Development Kit (NDK). Numerical results show that recent smartphones take less than 1 s to estimate the speed for each road segment in a network of 10 000 links from speed measurements at 1000 links. The proposed decentralized architecture significantly reduces the overhead of the communication network and paves the way for new cooperative traffic applications and operations.

Cell migration is a fundamental process for the development and maintenance of all multicellular organisms. Accurate cell tracking may lead to better interpretations of long-term cell behaviours. This paper describes an automated system to track multiple cells from experimental phase contrast images, which includes image registration, lumen segmentation, cell candidate detection, and multiple hypothesis Kalman filtering. We incorporate biological knowledge to associate the new observations to existing tracks. We apply our methodology to the problem of tracking endothelial cells in 3D angiogenic vessels. Numerical results indicate that our method associates cells more accurately compared to standard methods for cll association and tracking.

Multiscale (multiresolution) graphical models have gained widespread popularity in recent years, since they enjoy rich modeling power as well as efficient inference procedures. Existing approaches to learning multiscale graphical models often leverage the framework of penalized likelihood, and therefore suffer from the issue of regularization selection. In this paper, we propose a novel method to learn multiscale graphical models from the Bayesian perspective. More specifically, the regularization parameters are treated as random variables that follow Gamma distributions. We then derive an efficient variational Bayes algorithm to learn the model, and further demonstrate the advantages of the proposed method through numerical experiments.

To investigate the intraerythrocytic developmental cycle of Plasmodium falciparum, time-series gene expression data is commonly measured of infected red blood cells. However, the observed data are usually blurred due to cell asynchrony during experiments. In this paper, the effect of cell asynchrony is investigated by conducting numerical experiments. The simulation results suggest that cell asynchrony has varying effects on different intrinsic expression patterns. Specifically, the intrinsic patterns with high expression around the late life stage are more likely to be affected by cell asynchrony. It is also investigated how the effect of cell asynchrony depends on the experimental conditions. From this analysis, the burst rate r% in infection period and the standard deviation σ of growth rate are identified to have a strong impact on the blurring due to cell asynchrony. Consequently, it is important to measure these two parameters during biological experiments in order to deblur time-series gene expression data.

With advancements in sensor technologies, intelligent transportation systems can collect traffic data with high spatial and temporal resolution. However, the size of the networks combined with the huge volume of the data puts serious constraints on system resources. Low-dimensional models can help ease these constraints by providing compressed representations for the networks. In this paper, we analyze the reconstruction efficiency of several low-dimensional models for large and diverse networks. The compression performed by low-dimensional models is lossy in nature. To address this issue, we propose a near-lossless compression method for traffic data by applying the principle of lossy plus residual coding. To this end, we first develop a low-dimensional model of the network. We then apply Huffman coding (HC) in the residual layer. The resultant algorithm guarantees that the maximum reconstruction error will remain below a desired tolerance limit. For analysis, we consider a large and heterogeneous test network comprising of more than 18 000 road segments. The results show that the proposed method can efficiently compress data obtained from a large and diverse road network, while maintaining the upper bound on the reconstruction error.

Advanced sensing and surveillance technologies often collect traffic information with high temporal and spatial resolutions. The volume of the collected data severely limits the scalability of online traffic operations. To overcome this issue, we propose a low-dimensional network representation where only a subset of road segments is explicitly monitored. Traffic information for the subset of roads is then used to estimate and predict conditions of the entire network. Numerical results show that such approach provides 10 times faster prediction at a loss of performance of 3% and 1% for 5- and 30-min prediction horizons, respectively.

In this work we present a humanoid robot (Nao) that provides real-time sociofeedback to participants taking part in two-person dialogs. The sociofeedback system quantifies speech mannerism and social behavior of participants in an ongoing conversation, determines whether feedback is required, and delivers feedback through Nao. For example, Nao alarms the speaker(s) when the voice is too high or too low, or when the conversation is not proceeding well due to disagreements or numerous interruptions. In this study, participants are asked to engage in two-person conversations while the Nao robot acts as mediator. They then assess the received sociofeedback with respect to various aspects, including its content, appropriateness, and timing. Participants also evaluate their overall perception of Nao as social mediator via the Godspeed questionnaire.

Traffic forecasting is increasingly taking on an important role in many intelligent transportation systems (ITS) applications. However, prediction is typically performed for individual road segments and prediction horizons. In this study, we focus on the problem of collective prediction for multiple road segments and prediction-horizons. To this end, we develop various matrix and tensor based models by applying partial least squares (PLS), higher order partial least squares (HO-PLS) and N-way partial least squares (N-PLS). These models can simultaneously forecast traffic conditions for multiple road segments and prediction-horizons. Moreover, they can also perform the task of feature selection efficiently. We analyze the performance of these models by performing multi-horizon prediction for an urban subnetwork in Singapore.

With the rise in the population of the world's cities, understanding the dynamics of commuters' transportation patterns has become crucial in the planning and management of urban facilities and services. In this study, we analyze how commuter patterns change during different time instances such as between weekdays and weekends. To this end, we propose two data mining techniques, namely Common Orthogonal Basis Extraction (COBE), and Joint and Individual Variation Explained (JIVE) for Integrated Analysis of Multiple Data Types and apply them to smart card data available for passengers in Singapore. We also discuss the issues of model selection and interpretability of these methods. The joint and individual patterns can help transportation companies optimize their resources in light of changes in commuter mobility behavior.

Compressed Sensing (CS) provides a rich mathematical framework to efficiently acquire a sparse signal from few non-adaptive measurements. In radar imaging, most scenes are sparse and CS can be successfully applied for efficiently acquiring the target scene. Although the use of CS in radar is advantageous in many aspects, a higher noise in the received signal makes the output of CS unreliable. We propose a framework based on CS and matched filtering to improve the performance of CS particularly in high noise scenarios. We realize this framework by CS on chirp signal and discuss some limitations associated with it. Numerical experiments confirm a substantial performance improvement using the proposed framework compared to conventional CS reconstruction.

Meditation is a fascinating topic that is still relatively poorly understood. To investigate its physiological traits, electroencephalograms (EEG) were recorded during meditation sessions. In a recent study, paroxysmal gamma waves (PGWs) have been discovered in EEG of meditators practicing Bhramari Pranayama (BhPr). In this paper, the synchrony between those PGWs is investigated, revealing functional connectivity patterns in the brain during BhPr. Specifically, the method “Stochastic Event Synchrony” (SES) is applied to pairs of PGW sequences in order to assess their synchrony. From those pairwise synchrony measures, large-scale functional connectivity patterns are extracted. Three subjects possessing different levels of expertise in BhPr are considered. Strong synchrony can be observed in the temporal lobes for all three subjects, in addition to long-range interhemispheric connections. Consistent connectivity patterns are present for exhalation periods of BhPr, while those patterns are substantially less stationary for inhalation periods. Interestingly, the synchrony seems to increase gradually during the meditation sessions. Moreover, the distribution of synchrony values seems to depend on the level of expertise in practicing BhPr: the higher the expertise, the more concentrated the intensity values.

Measurement faults in control systems may result in permanent damages to the system components. Therefore, sensor validation is essential before the measurements are used for any system reconfiguration. In this paper, a statistical approach for sensor fault identification is proposed. Specifically, the potential sensor fault is assumed to be an additive bias term in the measurement model. The problem of fault identification is formulated as a least-squares optimization problem with an ℓ1penalty on the bias term. An algorithm is further introduced to determine the regularization parameter automatically. Experimental results show that the proposed method can accurately detect multiple sensor failures from noisy measurements.

Traffic prediction lies at the core of many intelligent transport systems (ITS). Commonly deployed prediction methods such as support vector regression and neural networks achieve good performance by explicitly predicting the traffic variables (e.g., traffic speed or volume) at each road segment in the network. For large traffic networks, predicting traffic variable at each road segment may be unwieldy, especially in the setting of real-time prediction. To tackle this problem, we propose an alternative approach in this paper. We first generate low-dimensional representation of the network, leveraging on the column-based (CX) decomposition of matrices. The low-dimensional model represents the large network in terms of a small subset of road segments. The future state of the low-dimensional network is predicted by standard procedures, i.e., support vector regression. The future state of the entire network is then inferred by extrapolating the predictions of the subnetwork, using the CX decomposition. Numerical results for a large-scale road network in Singapore demonstrate the efficiency and accuracy of the proposed algorithm.

To assess the risk of extreme events such as hurricanes and floods, it is crucial to develop accurate extreme-value statistical models. Extreme events often display heterogeneity, varying continuously with a number of covariates. Previous studies have suggested that models considering covariate effects lead to reliable estimates of extreme value distributions. In this paper, we develop a novel model to incorporate the effects of multiple covariates. Specifically, we analyze as an example the extreme sea states in the Gulf of Mexico, where the distribution of extreme wave heights changes systematically with location and wind direction. The block maxima at each location and sector of wind direction are assumed to follow the Generalized Extreme Value (GEV) distribution. The GEV parameters are coupled across the spatio-directional domain through a graphical model, particularly, a multidimensional thin-membrane model. Efficient learning and inference algorithms are then developed based on the special characteristics of the thin-membrane model. Numerical results for both synthetic and real data indicate that the proposed model can accurately describe marginal behavior of extreme events.

Graphical models are powerful tools to describe complex systems. Especially sparse graphical models are currently en vogue, as they allow us to infer network structure from multiple time series (e.g., functional brain networks from multichannel electroencephalograms). So far, most of the literature deals with stationary time series, whereas real-life time series often exhibit non-stationarity. In this paper, techniques are proposed to infer graphical models from piecewise stationary time series; first change point are detected in the time series, and then graphical models are inferred for each stationary segment. Specifically, a low-complexity algorithm based on Pruned Exact Linear Time method is proposed to identify change points. Copula Gaussian graphical models (with and without hidden variables) are then generated for each stationary segment. The crux of the proposed approach is that it determines the number and location of the change points as well as the graphical models in a fully automated manner. Results for both synthetic data and scalp electroencephalograms of epileptic seizure patients are provided to validate the model.

Assessing the risk of extreme events in a spatial domain, such as hurricanes, floods and droughts, presents unique significance in practice. Unfortunately, the existing extreme-value statistical models are typically not feasible for practical large-scale problems. Graphical models are capable of handling enormous number of variables, yet have not been explored in the realm of extreme-value analysis. To bridge the gap, an extreme-value graphical model is introduced in this paper, i.e., ensemble-of-trees of pairwise copulas (ETPC). In the proposed graphical model, extreme-value marginal distributions are stitched together by means of pairwise copulas, which in turn are the building blocks of the ensemble of trees. By exploiting this particular structure, novel efficient inference algorithms are derived that are applicable to large-scale statistical problems involving extreme values. It is proven that, under mild conditions, the ETPC model exhibits the favorable property of tail-dependence between an arbitrary pair of sites (variables), and therefore is reliable to capture the dependence between extremes at different sites. Real data results further demonstrate the advantages of the ETPC model.

Gaussian process (GP) regression is a nonparametric regression method that can be used to predict continuous quantities. Here, we show that the same technique can be applied to a class of phase imaging techniques based on measurements of intensity at multiple propagation distances, i.e. the transport of intensity equation (TIE). In this paper, we demonstrate how to apply GP regression to estimate the first intensity derivative along the direction of propagation and incorporate non-uniform propagation distance sampling. The low-frequency artifacts that often occur in phase recovery using traditional methods can be significantly suppressed by the proposed GP TIE method. The method is shown to be stable with moderate amounts of Gaussian noise. We validate the method experimentally by recovering the phase of human cheek cells in a bright field microscope and show better performance as compared to other TIE reconstruction methods.

We propose a novel statistical model to describe spatio-temporal extreme events. The model can be used to estimate extreme-value temporal pattern such as seasonality and trend, and further to predict the distribution of extreme events in the future. The basic idea is to explore graphical models to capture the highly structured dependencies among extreme events measured in time and space. More explicitly, we first assume the single observation at each location and time point follows a Generalized Extreme Value (GEV) distribution. The spatio-temporal dependencies are further encoded via graphical models imposed on the GEV parameters. We develop efficient learning and inference algorithms for the resulting non-Gaussian graphical model. Results of both synthetic and real data demonstrate the effectiveness of the proposed approach.

Neural simulation has been widely suggested as an alternative therapy for the treatment of medically-intractable seizures. Appropriate targeting of control stimuli at selected cortical locations may lead to seizure abatement. Neural population models describe the macroscopic neural activity that can be clinically recorded by an electroencephalogram (EEG). These models provide a safer way to develop and test the effect of such simulation strategies. In this study, a heterogeneously connected neural field model has been used which can replicate spatio-temporal patterns commonly observed in the EEG during generalized seizures. Seizure abatement has been formulated as an optimal control problem and the pseudospectral method has been used to develop stimuli with anti-ictogenic properties. The minimum energy optimal stimuli, developed in this study have been shown to abate seizures simulated from the model. It has been demonstrated that the control stimuli are spatially variant due to the underlying heterogeneity of the neural dynamics. This study provides a novel approach for designing optimal stimuli for seizure abatement while taking into account the heterogeneous dynamics of the human brain. It also raises the possibility of finding the appropriate set of cortical locations which may be stimulated to achieveP the anti-seizure effect.

A common problem with self-report quality-of-life questionnaires is missing data. Despite enormous care and effort to prevent it, some level of missing data is common and unavoidable. Missing data can have a detrimental impact on the data analysis. In this paper, a novel approach to imputing missing data in quality-of-life questionnaires is proposed, based on matrix and tensor decompositions. In order to illustrate and assess those methods, two datasets are considered: The first dataset contains the responses of 100 patients to a systemic lupus erythematosus-specific quality-of-life questionnaire; the other contains the responses of 43 patients to a rhino-conjunctivitis quality-of-life questionnaire. The two datasets contain almost no missing data, and for testing purposes, data entries are removed at random to have missing completely at random data. Several proportions of missing values are considered, and for each, the imputation error is assessed through k-fold cross validation. We also evaluate different imputation methods for missing at random and missing not at randomdata. The numerical results demonstrate that the proposed tensor factorization-based methods outperform standard methods in terms of root mean square error with at least 4% improvement, while the bias and variance are similar.

The ability to accurately predict traffic speed in a large and heterogeneous road network has many useful applications, such as route guidance and congestion avoidance. In principle, data-driven methods, such as support vector regression (SVR), can predict traffic with high accuracy because traffic tends to exhibit regular patterns over time. However, in practice, the prediction performance can significantly vary across the network and during different time periods. Insight into those spatiotemporal trends can improve the performance of intelligent transportation systems. Traditional prediction error measures, such as the mean absolute percentage error, provide information about the individual links in the network but do not capture global trends. We propose unsupervised learning methods, such as k-means clustering, principal component analysis, and self-organizing maps, to mine spatiotemporal performance trends at the network level and for individual links. We perform prediction for a large interconnected road network and for multiple prediction horizons with an SVR-based algorithm. We show the effectiveness of the proposed performance analysis methods by applying them to the prediction data of the SVR.

Gene expression level is measured on a population of cells in the microarray experiments. Due to the inevitable decay of cell synchrony, the intrinsic gene expression patterns are blurred in the observed time-series microarray data. Furthermore, slight differences in the experimental condition could be amplified as the experiments continue, and eventually considerably influence the cell synchrony, which dominates the way how the intrinsic expression patterns are blurred in the observed microarray data. Therefore, the decay of cell synchrony can cause variability between different experiments. In this paper, we demonstrate how to utilize computational synchronization approach to eliminate the variability caused by cell synchrony between different microarray experiments of Plasmodium falciparum. Specifically, The intrinsic expression patterns are respectively reconstructed for each microarray data set. The preliminary evaluations conducted on synthetic data suggest that computational synchronization could be a useful approach to improve the consistency of gene expression patterns over multiple experiments.

Steady state visually evoked potentials (SSVEPs), recorded from the central nervous system of humans using electroencephalography (EEG) in response to visual stimuli, have recently gained attention in cognitive and clinical neuroscience [1]. Here, we fuse EEG SSVEPs with recordings from functional magnetic resonance imaging data (fMRI) utilizing the millisecond temporal resolution of the former and the excellent spatial resolution of the latter. In particular, we propose a spatio-frequency EEG/fMRI fusion framework to recover the frequency content from the EEG and spatial information from the fMRI to enhance our understanding of SSVEPs. Notably, we consider fused analysis of fMRI and EEG data collected independently. We demonstrate that the proposed approach is a practical data-driven method to achieve spatio-frequency fusion of EEG and fMRI SSVEP responses.

Intelligent transport systems (ITS) require data with high spatial and temporal resolution for applications such as modeling, traffic management, prediction and route guidance. However, field data is usually quite sparse. This problem of missing data severely limits the effectiveness of ITS. Missing values are usually imputed by either using historical data of the road or current information from neighboring links. In most scenarios, information from some or all of neighboring links might not be available. Furthermore, historical data may also be incomplete. To overcome these issues, we propose methods which can construct low-dimensional representation of large and diverse networks, in presence of missing historical and neighboring data. We use these low-dimensional models to reconstruct data profiles for road segments, and impute missing values. To this end we use Fixed Point Continuation with Approximate SVD (FPCA) and Canonical Polyadic (CP) decomposition for incomplete tensors to solve the problem of missing data. We apply these methods to expressways and a large urban road network to assess their performance for different scenarios.

Copula Gaussian graphical models are capable of describing dependencies between a large number of heterogeneous variables. In this paper, low-complexity algorithms are proposed for learning copula Gaussian graphical models from discrete data. The proposed approach is Monte-Carlo expectation maximization: in the E-step, an efficient Gibbs sampler is applied, and in the M-step, the sparse graphical model is inferred by solving a penalized maximize likelihood problem. The regularization parameter is determined through the BINCO method proposed by Li et al. Numerical results for both synthetic and real data demonstrate the effectiveness of the proposed approach.

Meditation is a fascinating topic, yet has received limited attention in the neuroscience and signal processing community so far. A few studies have investigated electroencephalograms (EEG) recorded during meditation. Strong EEG activity has been observed in the left temporal lobe of meditators. Meditators exhibit more paroxysmal gamma waves (PGWs) in active regions of the brain. In this paper, a method is proposed to automatically detect PGWs from meditation EEG. The proposed algorithm is able to identify multiple sources in the brain that generate PGWs, and the sources associated with different types of PGWs can be distinguished. The effectiveness of the proposed method is assessed on 3 subjects possessing different degrees of expertise in practicing a yoga type meditation known as Bhramari Pranayama.

A novel near-lossless compression algorithm for multichannel electroencephalogram (MC-EEG) is proposed based on matrix/tensor decomposition models. MC-EEG is represented in suitable multiway (multidimensional) forms to efficiently exploit temporal and spatial correlations simultaneously. Several matrix/tensor decomposition models are analyzed in view of efficient decorrelation of the multiway forms of MC-EEG. A compression algorithm is built based on the principle of “lossy plus residual coding,” consisting of a matrix/tensor decomposition-based coder in the lossy layer followed by arithmetic coding in the residual layer. This approach guarantees a specifiable maximum absolute error between original and reconstructed signals. The compression algorithm is applied to three different scalp EEG datasets and an intracranial EEG dataset, each with different sampling rate and resolution. The proposed algorithm achieves attractive compression ratios compared to compressing individual channels separately. For similar compression ratios, the proposed algorithm achieves nearly fivefold lower average error compared to a similar wavelet-based volumetric MC-EEG compression algorithm.

In this paper, lossless and near-lossless compression algorithms for multichannel electroencephalogram signals (EEG) are presented based on image and volumetric coding. Multichannel EEG signals have significant correlation among spatially adjacent channels; moreover, EEG signals are also correlated across time. Suitable representations are proposed to utilize those correlations effectively. In particular, multichannel EEG is represented either in the form of image (matrix) or volumetric data (tensor), next a wavelet transform is applied to those EEG representations. The compression algorithms are designed following the principle of lossy plus residual coding, consisting of a wavelet-based lossy coding layer followed by arithmetic coding on the residual. Such approach guarantees a specifiable maximum error between original and reconstructed signals. The compression algorithms are applied to three different EEG datasets, each with different sampling rate and resolution. The proposed multichannel compression algorithms achieve attractive compression ratios compared to algorithms that compress individual channels separately.

This paper presents innovative collaborative filtering techniques to complete missing data in repeated medical questionnaires. The proposed techniques are based on the canonical polyadic (CP) decomposition (a.k.a. PARAFAC). Besides the standard CP decomposition, also a normalized decomposition is utilized. As an illustration, systemic lupus erythematosus-specific quality-of-life questionnaire is considered. Measures such as normalized root mean square error, bias and variance are used to assess the performance of the proposed tensor-based methods in comparison with other widely used approaches, such as mean substitution, regression imputations and k-nearest neighbor estimation. The numerical results demonstrate that the proposed methods provide significant improvement in comparison to popular methods. The best results are obtained for the normalized decomposition.

Various compression algorithms for multi-channel electroencephalograms (EEG) are proposed and compared. The multi-channel EEG is represented as a three-way tensor (or 3D volume) to exploit both spatial and temporal correlations efficiently. A general two-stage coding framework is developed for multi-channel EEG compression. In the first stage, we consider (i) wavelet-based volumetric coding; (ii) energy-based lossless compression of wavelet subbands; (iii) tensor decomposition based coding. In the second stage, the residual is quantized and coded. Through such two-stage approach, one can control the maximum error (worst-case distortion). Numerical results for a standard EEG data set show that tensor-based coding achieves lower worst-case error and comparable average error than the wavelet- and energy-based schemes.

Gaussian hidden variable graphical models are powerful tools to describe high-dimensional data; they capture dependencies between observed (Gaussian) variables by introducing a suitable number of hidden variables. However, such models are only applicable to Gaussian data. Moreover, they are sensitive to the choice of certain regularization parameters. In this paper, (1) copula Gaussian hidden variable graphical models are introduced, which extend Gaussian hidden variable graphical models to non-Gaussian data; (2) the sparsity pattern of the hidden variable graphical model is learned via stability selection, which leads to more stable results than cross-validation and other methods to select the regularization parameters. The proposed methods are validated on synthetic and real data.

Novel efficient algorithms are developed to infer the phase of a complex optical field from a sequence of intensity images taken at different defocus distances. The non-linear observation model is approximated by a linear model. The complex optical field is inferred by iterative Kalman smoothing in the Fourier domain: forward and backward sweeps of Kalman recursions are alternated, and in each such sweep, the approximate linear model is refined. By limiting the number of iterations, one can trade off accuracy vs. complexity. The complexity of each iteration in the proposed algorithm is in the order of N logN, where N is the number of pixels per image. The storage required scales linearly with N. In contrast, the complexity of existing phase inference algorithms scales with N3and the required storage with N2. The proposed algorithms may enable real-time estimation of optical fields from noisy intensity images.

A novel spatial model for extreme events is proposed. The model may for instance be used to describe the occurrence of catastrophic events such as earthquakes, floods, or hurricanes in certain regions; it may therefore be relevant for, e.g., weather forecasting, urban planning, and environmental assessment. The model is derived from the following ideas: The above-threshold values at each location are assumed to follow a generalized Pareto (GP) distribution. The GP parameters are coupled across space through Markov random fields, in particular, thin-membrane models. The latter are inferred through an empirical Bayes approach. Numerical results are presented for synthetic and real data (related to hurricanes in the Gulf of Mexico).

Stochastic event synchrony (SES) is a recently proposed family of similarity measures. First, "events" are extracted from the given signals; next, one tries to align events across the different time series. The better the alignment, the more similar the N time series are considered to be. The similarity measures quantify the reliability of the events (the fraction of "nonaligned" events) and the timing precision. So far, SES has been developed for pairs of one-dimensional (Part I) and multidimensional (Part II) point processes. In this letter (Part III), SES is extended from pairs of signals to N > 2 signals. The alignment and SES parameters are again determined through statistical inference, more specifically, by alternating two steps: (1) estimating the SES parameters from a given alignment and (2), with the resulting estimates, refining the alignment. The SES parameters are computed by maximum a posteriori (MAP) estimation (step 1), in analogy to the pairwise case. The alignment (step 2) is solved by linear integer programming. In order to test the robustness and reliability of the proposed N-variate SES method, it is first applied to synthetic data. We show that N-variate SES results in more reliable estimates than bivariate SES. Next N-variate SES is applied to two problems in neuroscience: to quantify the firing reliability of Morris-Lecar neurons and to detect anomalies in EEG synchrony of patients with mild cognitive impairment. Those problems were also considered in Parts I and II, respectively. In both cases, the N-variate SES approach yields a more detailed analysis.

We have developed a computational approach which predicts the protein kinases that may regulate the transition between the blood developmental stages of Plasmodium falciparum (P. falciparum). To improve the accuracy of our pre diction, synchronized gene expression levels are reconstructed from the observed microarray data generated by the ensembles of non-synchronized cells. Peaks in annotated protein kinase transcript levels are hypothesized to directly correlate with the period when the encoded protein kinases function temporally. Therefore, protein kinases, which putatively regulate a given developmental stage transition, are identified by their peak in synchronized gene expression levels. By analyzing publicly available microarray data set, a few protein kinases are considered to be strongly associated with developmental stage transition. Two of these (PF13_0211, PFB0815w) have recently been implicated in the schizont to ring transition [1], [2]. Another one of these identified (MAL7P1.144) has been found to influence erythrocyte membrane in both trophozoite and schizont [3]. Overall, these results suggest that further functional analysis of the other protein kinases we have predicted may reveal new insights into P. falciparum blood stage development.

Decision algorithms are developed that use periods of intracranial non-seizure (interictal) EEG to localize epileptogenic networks. Depth and surface recordings are considered from 5 and 6 patients respectively. The proposed algorithms combine spectral and multivariate statistics in a decision-theoretic framework to automatically delineate the seizure onset area. In the case of depth recordings, we apply standard binary classification algorithms, including linear and quadratic discriminative analysis. For the surface recordings, novel decision algorithms are developed, based upon graphical models. The outcomes from the algorithms for both depth and surface recordings are in good agreement with the determination of the seizure focus by clinicians from ictal EEG. In the long term, the proposed approach may lead to shorter hospitalization of intractable-epilepsy patients, since it does not rely on ictal EEG.

Compression schemes for EEG signals are developed based on matrix and tensor decomposition. Various ways to arrange EEG signals into matrices and tensors are explored, and several matrix and tensor decomposition schemes are applied, including SVD, CUR, PARAFAC, the Tucker decomposition, and recent random fiber selection approaches. Rate-distortion curves for the proposed matrix and tensor-based EEG compression schemes are computed. It shown that PARAFAC has the best compression performance in this context.

In this paper, we consider the problem of quantifying synchrony between multiple simultaneously recorded electroencephalographic signals. These signals exhibit nonlinear dependencies and non-Gaussian statistics. A copula based approach is presented to model the joint statistics. We then consider the application of copula derived synchrony measures for early diagnosis of Alzheimer's disease. Results on real data are presented.

An explicit code construction for using low-density lattice codes (LDLC) on the constrained power AWGN channel is given. LDLC lattices can be decoded in high dimension, so that the code relies on the Euclidean distance between codepoints. A sublattice of the coding lattice is used for code shaping. Lattice codes are designed using the continuous approximation, which allows separating the contribution of the shaping region and coding lattice to the total transmit power. Shaping and lattice decoding are both performed using a belief-propagation decoding algorithm. At a rate of 3 bits per dimension, a dimension 100 code which is 3.6 dB from the sphere bound is found.

We present a novel approach to quantify the statistical interdependence of two time series, referred to as stochastic event synchrony (SES). The first step is to extract the two given time series. The next step is to try to align events from one time series with events from the other. The better the alignment the more similar the two series are considered to be. More precisely, the similarity is quantified by the following parameters: time delay, variance of the time jitter, fraction of noncoincident events, and average similarity of the aligned events. The pairwise alignment and SES parameters are determined by statistical inference. In particular, the SES parameters are computed by maximum a posteriori (MAP) estimation, and the pairwise alignment is obtained by applying the max product algorithm. This letter deals with one-dimensional point processes; the extension to multidimensional point processes is considered in a companion letter in this issue. By analyzing surrogate data, we demonstrate that SES is able to quantify both timing precision and event reliability more robustly than classical measures can. As an illustration, neuronal spike data generated by Morris-Lecar neuron model are considered.

Stochastic event synchrony is a technique to quantify the similarity of pairs of signals. First, events are extracted from the two given time series. Next, one tries to align events from one time series with events from the other. The better the alignment, the more similar the two time series are considered to be. In Part I, the companion letter in this issue, one-dimensional events are considered; this letter concerns multidimensional events. Although the basic idea is similar, the extension to multidimensional point processes involves a significantly more difficult combinatorial problem and therefore is nontrivial. Also in the multidimensional case, the problem of jointly computing the pairwise alignment and SES parameters is cast as a statistical inference problem. This problem is solved by coordinate descent, more specifically, by alternating the following two steps: (1) estimate the SES parameters from a given pairwise alignment; (2) with the resulting estimates, refine the pairwise alignment. The SES parameters are computed by maximum a posteriori (MAP) estimation (step 1), in analogy to the one-dimensional case. The pairwise alignment (step 2) can no longer be obtained through dynamic programming, since the state space becomes too large. Instead it is determined by applying the max-product algorithm on a cyclic graphical model. In order to test the robustness and reliability of the SES method, it is first applied to surrogate data. Next, it is applied to detect anomalies in EEG synchrony of mild cognitive impairment (MCI) patients. Numerical results suggest that SES is significantly more sensitive to perturbations in EEG synchrony than a large variety of classical synchrony measures.

A novel approach to measure the interdependence of time series is proposed, based on the alignment ("matching") of their Huang-Hilbert spectra. The method consists of three steps: first, empirical modes are extracted from the signals; those functions carry non-linear and non-stationary components in frequency limited bands. Second, the empirical modes are Hilbert transformed, resulting in very sharply localized ridges in the time- frequency plane; the obtained time-frequency representations are known as Huang-Hilbert spectra. At last, the latter are pairwise aligned by means of the stochastic-event synchrony method (SES), a recently proposed procedure to match pairs of multi-dimensional point processes. The level of similarity of two Huang-Hilbert spectra is quantified by three parameters: timing and frequency jitter of coincident ridges, and fraction of non-coincident ridges. The proposed method is used to detect steady-state visually evoked potentials (SSVEP) in electroencephalography (EEG) signals; numerical results indicate that the method is vastly more sensitive to SSVEP than classical synchrony measures, and therefore, it may prove to be useful in applications such as brain-computer interfaces. Although the paper mostly deals with EEG, the presented synchrony measure may also be applied to other kinds of time series.

A belief-propagation decoder for low-density lattice codes, which represents messages explicitly as a mixture of Gaussians functions, is given. In order to prevent the number of functions from growing as the decoder iterations progress, a method for reducing the number of Gaussians at each step is given. A squared distance metric is used, which is shown to be a lower bound on the divergence. For an unconstrained power system, comparisons are made with a quantized implementation. For a dimension 100 lattice, a loss of about 0.2 dB was found; for dimension 1000 and 10000 lattices, the difference in error rate was indistinguishable. The memory required to store the messages is substantially superior to the quantized implementation.

Prior work on the computation of information rates of channels with memory is extended to continuous state spaces by means of sequential Monte-Carlo integration (ldquoparticle filteringrdquo).

An algorithm is proposed that can be used (i) to verify the validity of the asymptotic Cramer-Rao bound (CRB) for a finite number of observations; (ii) to improve the CRB. It is derived from the fibre bundle theory of local exponential families developed by Amari and by Barndorff-Nielsen et al. The algorithm is applicable to any regular estimation problem. As an illustration, the problem of estimating the parameters of an AR model with noisy observations is considered.

A novel measure to quantify the synchrony between two sparse binary strings is proposed, referred to as "stochastic event synchrony" (SES). It is computed by performing inference in a probabilistic model. SES can amongst other be used to detect synchrony in neural signals, in particular, spike trains (obtained from electrophysiological recordings) and EEG signals. It is demonstrated how SES can quantify the firing reliability of a neuron. It is also shown how SES can be used as a feature to detect Alzheimer's disease based on EEG signals.

In this paper, it is shown how (naive and structured) variational algorithms may be derived from a factor graph by mechanically applying generic message computation rules; in this way, one can bypass error-prone variational calculus. In prior work by Bishop et al., Xing et al., and Geiger, directed and undirected graphical models have been used for this purpose. The factor graph notation amounts to simpler generic variational message computation rules; by means of factor graphs, variational methods can straightforwardly be compared to and combined with various other message-passing inference algorithms, e.g., Kalman filters and smoothers, iterated conditional modes, expectation maximization (EM), gradient methods, and particle filters. Some of those combinations have been explored in the literature, others seem to be new. Generic message computation rules for such combinations are formulated.

The convergence and stability properties of several message-passing estimation algorithms are investigated, i.e., (generalized) expectation maximization, gradient methods, and coordinate ascent/descent. Results are presented for cycle-free and cyclic factor graphs.

A novel approach to measure the interdependence of two time series is proposed, referred to as “stochastic event synchrony” (SES); it quantifies the alignment of two point processes by means of the following parameters: time delay, variance of the timing jitter, fraction of “spurious” events, and average similarity of events. SES may be applied to generic one-dimensional and multi-dimensional point pro- cesses, however, the paper mainly focusses on point processes in time-frequency domain. The average event similarity is in that case described by two parameters: the average frequency offset between events in the time-frequency plane, and the variance of the frequency offset (“frequency jitter”); SES then consists of five pa- rameters in total. Those parameters quantify the synchrony of oscillatory events, and hence, they provide an alternative to existing synchrony measures that quan- tify amplitude or phase synchrony. The pairwise alignment of point processes is cast as a statistical inference problem, which is solved by applying the max- product algorithm on a graphical model. The SES parameters are determined from the resulting pairwise alignment by maximum a posteriori (MAP) estimation. The proposed interdependence measure is applied to the problem of detecting anoma- lies in EEG synchrony of Mild Cognitive Impairment (MCI) patients; the results indicate that SES significantly improves the sensitivity of EEG in detecting MCI.

The message-passing approach to model-based signal processing is developed with a focus on Gaussian message passing in linear state-space models, which includes recursive least squares, linear minimum-mean-squared-error estimation, and Kalman filtering algorithms. Tabulated message computation rules for the building blocks of linear models allow us to compose a variety of such algorithms without additional derivations or computations. Beyond the Gaussian case, it is emphasized that the message-passing approach encourages us to mix and match different algorithmic techniques, which is exemplified by two different approaches—steepest descent and expectation maximization—to message passing through a multiplier node.

A numerical algorithm is proposed to compute Cramer-Rao-type bounds. The Cramer-Rao-type bounds are derived from information matrices of marginals of the joint pdf of the system at hand. The key ingredient is message-passing on a factor graph of the system. The method can be applied to a wide class of estimation problems. As an illustration, the problem of estimating the parameters of an AR model is considered

It is shown how particle methods can be viewed as message passing on factor graphs. In this setting, particle methods can readily be combined with other message-passing techniques such as the sum-product and max-product algorithm, expectation maximization, iterative conditional modes, steepest descent, Kaiman filters, etc. Generic message computation rules for particle-based representations of sum-product messages are formulated. Various existing particle methods are described as instances of those generic rules, i.e., Gibbs sampling, importance sampling, Markov-chain Monte Carlo methods (MCMC), particle filtering, and simulated annealing

It has been observed that a linear-feedback shift-register (LFSR) sequence can be synchronized by feeding the modulated sequence into a "soft" (or "analog") version of the LFSR. In this correspondence, the "soft LFSR" is derived as forward-only message passing in the corresponding factor graph. A continous-time analog (suitable for realization as a clockless electronic circuit) is then given of both the LFSR and the soft LFSR. A connection is thus established between statistical state estimation and the phenomenon of entrainment of dynamical systems, which opens the prospect of deriving dynamical systems (such as electronic circuits) with strong entrainment capabilities from more powerful message passing algorithms

An efficient message-passing algorithm for computing the Bayesian Cramer-Rao bound (BCRB) for general estimation problems is presented. The BCRB is a lower bound on the mean squared estimation error. The algorithm operates on a cycle-free factor graph of the system at hand. It can be applied to estimation in (1) general state-space models; (2) coupled state-space models and other systems that are most naturally represented by cyclic factor graphs; (3) coded systems

Based on prior work by Eckford, it is shown how expectation maximization (EM) may be viewed, and used, as a message passing algorithm in factor graphs

It is shown how steepest descent (or steepest ascent) may be viewed as a message passing algorithm with "local" message update rules. For example, the well-known backpropagation algorithm for the training of feedforward neural networks may be viewed as message passing on a factor graph. The factor graph approach with its emphasis on "local" computations makes it easy to combine steepest descent with other message passing algorithms such as the sum/max-product algorithms, expectation maximization, Kalman filtering/smoothing, and particle filters. As an example, parameter estimation in a state space model is considered. For this example, it is shown how steepest descent can be used for the maximization step in expectation maximization.

The problem of phase estimation in a "turbo receiver" is considered for two different channel models. Several message passing algorithms for phase estimation are derived from the factor graph of the channel models: (1) straight sum-product, applied to a quantized phase model; (2) LMS-type gradient methods; (3) a particle filter. All considered algorithms are suitable for use in a "turbo receiver" with joint iterative decoding and phase estimation.

Prior work on the computation of information rates of channels with memory is extended to continuous state spaces by means of sample-based numerical integration ("particle filtering"). In this paper, the problem of computing the information rate between the input and the output process of a time-variant discrete-time channel with memory and ergodic stochastic process is analyzed and the methods are extended to continuous state spaces.