Unseen Time Research Articles

Time series anomaly detection using generative adversarial network discriminators and density estimation for infrastructure systems

Efficient data-driven defect detection techniques are crucial for maintaining service quality and providing early warnings for infrastructure systems. To this end, we proposed an effective unsupervised anomaly detection framework (DEGAN) using Generative Adversarial Networks (GANs). The framework relies solely on normal time series data as input to train well-configured discriminators into standalone anomaly predictors by leveraging repeatedly collected data from an infrastructure system. Expected normal patterns in data are identified by generators, and well-configured discriminators are extracted to evaluate anomalies in unseen time series. Kernel density estimation (KDE) is coupled with discriminators for probabilistic anomaly detection. Through a Class I railroad track case study, we evaluated the performance of a convolutional DEGAN in detecting anomalies identified by operators, achieving recall and precision of 80% and 86%, respectively. We also investigated the influence of GAN architectures and parameters, model validation scheme (supervised vs. unsupervised), clustering, and the KDE parameters.

A machine learning-based algorithm for automated detection of frequency-based events in recorded time series of sensor data

Automated event detection methods, vital for monitoring technical systems via sensor data, are highly sought after in the automotive industry for tracing events in time series data. For assessing the active vehicle safety systems, a diverse range of driving scenarios is conducted. These scenarios involve the recording of the vehicle's behavior using external sensors, enabling the evaluation of operational performance. In such setting, automated detection methods not only accelerate but also standardize and objectify the evaluation by avoiding subjective, human-based appraisals in the data inspection. This study introduces a novel method for identifying frequency-based events in time series data. To this aim, the time series data is mapped to representations in the time-frequency domain, known as scalograms. After filtering scalograms to enhance relevant parts of the signal, an object detection model is trained to detect the desired event objects in the scalograms. By leveraging the accuracy of object detection networks in localizing objects, precise time intervals of events are identified. For the analysis of unseen time series data, events can be detected in their scalograms with the trained object detection model and are thereafter mapped back to the time series data to mark the corresponding time interval. The algorithm, evaluated on unseen datasets, achieves a precision rate of 0.97 in event detection, providing sharp time interval boundaries whose accurate indication by human visual inspection is challenging. Incorporating this method into the vehicle development process enhances the accuracy and reliability of event detection, which holds major importance for rapid testing analysis.

An encoder-decoder ConvLSTM surrogate model for simulating geological CO2 sequestration with dynamic well controls

In Geological Carbon Sequestration (GCS), effectively managing the project requires predicting state variables such as pressure and saturation. However, numerical simulation of multiphase flow in subsurface porous media involves solving large linear algebra systems, resulting in a substantial computational burden. This can make it impractical for real-time history matching or optimization. Meanwhile, deep learning-based surrogate models are emerging as fast and accurate approximators. This study delves into the application of the Encoder-Decoder Convolutional Long Short-Term Memory (ED-ConvLSTM) neural network for predicting the complex evolution of state variables under dynamic CO2 injection schemes. Inception blocks enhanced with light-weighted attention modules are introduced in the encoder to extract high-dimensional input features. ConvLSTM is employed to propagate spatial temporal information in the low-dimensional latent space. Further, progressive upsampling blocks are used to reconstruct the latent features for the desired output. Instead of taking discrete time steps as an input feature, the proposed network captures the dynamic dependencies with the inherent ConvLSTM cell. The network has access to data at only portion of the initial time steps during training stage, while it is used to predict the state variables at unseen time steps during testing stage. Results show that the network can produce excellent predictions for both pressure and saturation, even at unseen future time steps. The remarkable generalizability to different geological permeability fields is also evaluated. ED-ConvLSTM outperforms the standard U-Net by far, especially when predicting beyond the training time period. These numerical experiments demonstrate the advantages of ED-ConvLSTM in terms of prediction accuracy, extrapolability and generalizability. This study highlights the importance of incorporating recurrent connections into the deep neural networks for simulating time-dependent multiphase flow problems. The proposed methodology has great potential in GCS surrogate modeling and offers a possible approach for real-time optimization of CO2 injection.

Residual-based physics-informed transfer learning: A hybrid method for accelerating long-term CFD simulations via deep learning

While a big wave of artificial intelligence (AI) has propagated to the field of computational fluid dynamics (CFD) acceleration studies, recent research has highlighted that the development of AI techniques that reconciles the following goals remains our primary task: (1) accurate prediction of unseen (future) time series in long-term CFD simulations (2) acceleration of simulations (3) an acceptable amount of training data and time (4) within a multiple PDEs condition. In this study, we propose a residual-based physics-informed transfer learning (RePIT) strategy to achieve these four objectives using ML-CFD hybrid computation. Our hypothesis is that long-term CFD simulation is feasible with the hybrid method where CFD and AI alternately calculate time series while monitoring the first principle's residuals. The feasibility of RePIT strategy was verified through a CFD case study on natural convection. In a single training approach, a residual scale change occurred around 100th timestep, resulting in predicted time series exhibiting non-physical patterns as well as a significant deviations from the ground truth. Conversely, RePIT strategy maintained the residuals within the defined range and demonstrated good accuracy throughout the entire simulation period. The maximum error from the ground truth was below 0.4 K for temperature and 0.024 m/s for x-axis velocity. Furthermore, the average time for 1 timestep by the ML-GPU and CFD-CPU calculations was 0.171 s and 0.015 s, respectively. Including the parameter-updating time, the simulation was accelerated by a factor of 1.9. In conclusion, our RePIT strategy is a promising technique to reduce the cost of CFD simulations in industry. However, more vigorous optimization and improvement studies are still necessary.

Pre-triage wait times for non-ambulance arrivals in the emergency department: A retrospective video audit

BackgroundPre-triage emergency department (ED) waiting times can be lengthy when presentation numbers are high. Queuing is random, affecting flow management and patient care. We investigated pre-triage wait times and barriers to triage access at an Australian ED. MethodsA reviewer conducted a retrospective audit of triage reception security video camera footage (February–March, 2020). The reviewer manually documented self-presenting patients’ wait-to-be-seen times and barriers to patient flow. ResultsThe audit identified three main topics: lengthy pre-triage wait times, pre-triage queuing and observed barriers to triage. Median pre-triage wait time was 12 min (IQR = 5–21; n = 141), with no apparent relationship between patients’ wait time and time of arrival. During peak or busy periods, multiple random queues formed at the triage reception area. Triage nurses could not concurrently triage and provide queue control during busy periods. ConclusionsUnrecorded pre-triage wait times may exceed 20 min. This unseen time may extend beyond the estimated post-triage wait times suggested by the Australasian Triage Scale (ATS). There was a degree of disorder in patient queueing, reducing effective door-to-triage productivity. Larger studies could determine these findings’ external replicability, with additional research addressing potential benefits of pre-triage queuing processes or a departmental concierge.

V2V: A Deep Learning Approach to Variable-to-Variable Selection and Translation for Multivariate Time-Varying Data.

We present V2V, a novel deep learning framework, as a general-purpose solution to the variable-to-variable (V2V) selection and translation problem for multivariate time-varying data (MTVD) analysis and visualization. V2V leverages a representation learning algorithm to identify transferable variables and utilizes Kullback-Leibler divergence to determine the source and target variables. It then uses a generative adversarial network (GAN) to learn the mapping from the source variable to the target variable via the adversarial, volumetric, and feature losses. V2V takes the pairs of time steps of the source and target variable as input for training, Once trained, it can infer unseen time steps of the target variable given the corresponding time steps of the source variable. Several multivariate time-varying data sets of different characteristics are used to demonstrate the effectiveness of V2V, both quantitatively and qualitatively. We compare V2V against histogram matching and two other deep learning solutions (Pix2Pix and CycleGAN).

EPTs-TL: A two-level approach for efficient event prediction in healthcare

Recently, the event prediction on time series (EPTs) was discussed as one of the important and interesting research trends that its usage is growing for taking proper decisions in the various sciences. In the real-world, time series event-based analysis can pose as one of the challenging prediction problems in healthcare, which have a direct impact and a key role in supporting health management. In this paper, an efficient approach of two-level (TL) is proposed to the EPTs problem in healthcare, which named EPTs-TL. At the first level, unseen time series data is predicted by using an enhanced hybrid model based on soft computing technology. Then, a new feature extraction-based method is proposed for fuzzy detection of future events in two-level. The EPTs -TL approach employed concepts of three components: weighting, fuzzy logic, and metaheuristics in two-level of the proposed approach. The empirical results demonstrate the excellent performance of the EPTs -TL approach in comparison to conventional prediction models in healthcare and medicine. Also, the proposed approach can be introduced as a strong tool to handle the complex and uncertain behaviors of time series, analyze unusual variations of those, forewarn the possible critical situations in the society, and fuzzy predict event in healthcare.

An autoencoder-based deep learning approach for clustering time series data

This paper introduces a two-stage deep learning-based methodology for clustering time series data. First, a novel technique is introduced to utilize the characteristics (e.g., volatility) of the given time series data in order to create labels and thus enable transformation of the problem from an unsupervised into a supervised learning. Second, an autoencoder-based deep learning model is built to model both known and hidden non-linear features of time series data. The paper reports a case study in which the selected financial and stock time series data of over 70 stock indices are clustered into distinct groups using the introduced two-stage procedure. The results show that the proposed methodology is capable of achieving 87.5% accuracy in clustering and predicting the labels for unseen time series data. The paper also reports an important finding in which it is observed that the performance of both techniques (i.e., autoencoder and Kmeans) are comparable. However, there are a few instances of time series data that are classified differently by the autoencoder-based methodology compared to the Kmeans algorithm. The results may indicate that the proposed deep learning-based approach is taking into account additional hidden features that might be overlooked by conventional Kmeans. The finding raises the question whether the explicit features of data should be analyzed for clustering or more advanced techniques such as deep learning need to be adapted by which hidden features and relationships are explored for clustering purposes.

A learning algorithm for time series based on statistical features

In this study, we propose a machine learning technique for time-series data which combines statistical features and neural networks. The proposed algorithm is tested on various time series like stock prices, astronomical light curve and currency exchange rates. An implementation of reconstruction of unseen time series based on encodings learned by the neural network from the training data is proposed and tested. The predicted time series based on statistical features of past values show that the trained models were able to capture well the structure of the time-series data.

Wrinkles in Time: Ageing Costume in Hindi Film

Today’s philosophy and practice of costume ageing, even in mainstream commercial Bollywood output, skews strongly towards an avowed ‘realism’. Consequently, accurate ageing and the subtle impressions of wear are valued in contrast to the ‘theatrical’ and ‘inauthentic’ ageing of most pre-1990s films (and some films still today). Designers argue that costume ageing has simply improved but this answer oversimplifies the complex narrative and organisational imperatives at stake. Older, more theatrical costume ageing, embedded within the melodramatic mode of expression, worked for its audience because of the explicit contrast it drew with costumes that were pristine. The distinction between new and aged costumes served many functions, among them the marking of vulnerable versus invulnerable bodies. Stars, dressed in new, unworn clothes, achieved their near mythic identifications in part because their costumes resisted the rigors of time and experience. In this past era, it was sufficient to pile on dirt and tear fabric to achieve effective ‘ageing’ as opposed to carefully mimicking how clothes actually age. This type of quick, crude ageing was both a consequence of—and a rationalisation for—scant time spent in costume ageing (and fabrication) in pre-production. New practices that strive for ‘realistic’ ageing thrive in expanded pre-production schedules. Alongside a resilient poetics of aged costume, ‘relaxed’ costumes lend texture to the film’s ‘lived world’. Now, the goal of ageing is to index the unseen time that characters have experienced outside the film’s temporal boundaries.

Personalized app recommendation using spatio-temporal app usage log

This paper proposes a probabilistic method to recommend apps appropriate to current time and location of a user. The proposed method regards an app as a distribution of topics discovered from a large number of app descriptions. A user preference is then modeled, using spatio-temporal app usage log of a user, as a topic distribution that is affected by time and location. Since time and location can be regarded as two continuous random variables that are independent of each other, the proposed method is in contrast to conventional methods in that the conventional methods are based on a limited number of discrete contexts and assume that locations are dependent on time. Therefore, the proposed method captures user-specific contexts and is robust even in unseen time and location. Our experiments show that the proposed method outperforms two baseline methods in NDCG, which implies that the proposed method is effective in personalized app recommendation.

Bayesian state space models for dynamic genetic network construction across multiple tissues.

Construction of gene-gene interaction networks and potential pathways is a challenging and important problem in genomic research for complex diseases while estimating the dynamic changes of the temporal correlations and non-stationarity are the keys in this process. In this paper, we develop dynamic state space models with hierarchical Bayesian settings to tackle this challenge for inferring the dynamic profiles and genetic networks associated with disease treatments. We treat both the stochastic transition matrix and the observation matrix time-variant and include temporal correlation structures in the covariance matrix estimations in the multivariate Bayesian state space models. The unevenly spaced short time courses with unseen time points are treated as hidden state variables. Hierarchical Bayesian approaches with various prior and hyper-prior models with Monte Carlo Markov Chain and Gibbs sampling algorithms are used to estimate the model parameters and the hidden state variables. We apply the proposed Hierarchical Bayesian state space models to multiple tissues (liver, skeletal muscle, and kidney) Affymetrix time course data sets following corticosteroid (CS) drug administration. Both simulation and real data analysis results show that the genomic changes over time and gene-gene interaction in response to CS treatment can be well captured by the proposed models. The proposed dynamic Hierarchical Bayesian state space modeling approaches could be expanded and applied to other large scale genomic data, such as next generation sequence (NGS) combined with real time and time varying electronic health record (EHR) for more comprehensive and robust systematic and network based analysis in order to transform big biomedical data into predictions and diagnostics for precision medicine and personalized healthcare with better decision making and patient outcomes.

Probabilistic anomaly detection in natural gas time series data

This paper introduces a probabilistic approach to anomaly detection, specifically in natural gas time series data. In the natural gas field, there are various types of anomalies, each of which is induced by a range of causes and sources. The causes of a set of anomalies are examined and categorized, and a Bayesian maximum likelihood classifier learns the temporal structures of known anomalies. Given previously unseen time series data, the system detects anomalies using a linear regression model with weather inputs, after which the anomalies are tested for false positives and classified using a Bayesian classifier. The method can also identify anomalies of an unknown origin. Thus, the likelihood of a data point being anomalous is given for anomalies of both known and unknown origins. This probabilistic anomaly detection method is tested on a reported natural gas consumption data set.

Balancing target acquisition and target tracking in existing resource constrained stationary camera systems

Safety security concerns drive the need to visually monitor large, complex environments (such as college campuses and office parks). Operators use networks of cameras to surveil areas in a dual fashion: first, identify events or targets of interest and second, monitor newly identified events and targets until a determination of subsequent action is made. An intelligent camera motion algorithm can assist operators with surveilling large areas by leveraging the region of interest configuration. This article compares the usefulness of camera motion planning by minimizing the time a space goes unseen (time based) against minimizing the size of contiguous unviewable space (space based). In addition, an autonomous, multi-camera, cooperative surveillance algorithm that operates a network of pan---tilt---zoom cameras to monitor a region of interest is presented. Coverage is accomplished via a novel algorithm that combines both time and space-based heuristics without an exhaustive search for the optimal parameters. This approach not only cooperatively manages camera coverage, but introduces a mechanism for balancing resource assignment between coverage to acquire new targets and tracking of previously discovered targets. Other benefits include a user-interface mechanism and occlusion awareness for expansion to cluttered urban environments. Experimental results are presented and compared for target acquisition and tracking in a typical urban campus environment. Results show that average linear uncovered length (ALUL) better predicts target acquisition performance. In addition, the use of ALUL for selection of camera panning angles improves target tracking and surveillance performance for both static and dynamic surveillance systems.

Bayesian State Space Models for Inferring and Predicting Temporal Gene Expression Profiles

Prediction of gene dynamic behavior is a challenging and important problem in genomic research while estimating the temporal correlations and non-stationarity are the keys in this process. Unfortunately, most existing techniques used for the inclusion of the temporal correlations treat the time course as evenly distributed time intervals and use stationary models with time-invariant settings. This is an assumption that is often violated in microarray time course data since the time course expression data are at unequal time points, where the difference in sampling times varies from minutes to days. Furthermore, the unevenly spaced short time courses with sudden changes make the prediction of genetic dynamics difficult. In this paper, we develop two types of Bayesian state space models to tackle this challenge for inferring and predicting the gene expression profiles associated with diseases. In the univariate time-varying Bayesian state space models we treat both the stochastic transition matrix and the observation matrix time-variant with linear setting and point out that this can easily be extended to nonlinear setting. In the multivariate Bayesian state space model we include temporal correlation structures in the covariance matrix estimations. In both models, the unevenly spaced short time courses with unseen time points are treated as hidden state variables. Bayesian approaches with various prior and hyper-prior models with MCMC algorithms are used to estimate the model parameters and hidden variables. We apply our models to multiple tissue polygenetic affymetrix data sets. Results show that the predictions of the genomic dynamic behavior can be well captured by the proposed models.