Future Time Steps Research Articles

VAM: An End-to-End Simulator for Time Series Regression and Temporal Link Prediction in Social Media Networks

We present a machine-learning-driven end-to-end simulator, called the <i>Volume-Audience-Match</i> (VAM) simulator. VAM&#x2019;s purpose is to simulate future phenomena related to various topics of discussion in social media networks. We focus our attention on the social media platform, Twitter, due to its abundant use in today&#x2019;s world. VAM was applied to do time series forecasting to predict the future: 1) number of total activities; 2) number of active old users; and 3) number of newly active users over the span of 24 hours from the start time of prediction. VAM then used these macroscopic volume predictions (VPs) to perform user link predictions. A user&#x2013;user edge was assigned to each of the activities in the 24 future time steps. We report that VAM outperformed multiple baseline models in the time series task, which were the auto-regressive integrated moving average (ARIMA), auto regressive moving average (ARMA), auto regressive (AR), moving average (MA), Persistence Baseline, and state-of-the-art tNodeEmbed models. Furthermore, we show that VAM outperformed the Persistence Baseline and tNodeEmbed models used for the user-assignment tasks. Finally, it is also shown that using Reddit activity data improves prediction accuracy.

In-Excess Attenuation Detection Using Satellite Link Channel Measurements at Ka- and Q-Bands With Deep-Learning Architectures

In this paper a robust process to detect the existence of in excess attenuation values in the next future time-steps based on measured data from previous time-steps is provided for satellite channel measurements campaign at frequencies above 10 GHz. The straightforward process of anomaly detection is applied, where ’normal’ data, i.e. attenuation values lower than a threshold, are used to learn their predictive error probability. Afterwards a sequence is detected as ’normal’ or ’abnormal’ based on the Mahalanobis distance of its errors from the learnt distribution. Deep learning architectures, such as encoder-decoder and traditional paradigms are investigated and tested for different regions (Greece and U.K.) and frequency bands (Q and Ka downlinks). The method’s performance is very promising and the rainfall rate information seems valuable and adequate for attenuation’s detection. Finally, under a ’plug-and-play’ logic, the trained detectors in Greece exhibit high performance also in U.K. data showing the robustness of the proposed detection ability.

HiC4D: forecasting spatiotemporal Hi-C data with residual ConvLSTM.

The Hi-C experiments have been extensively used for the studies of genomic structures. In the last few years, spatiotemporal Hi-C has largely contributed to the investigation of genome dynamic reorganization. However, computationally modeling and forecasting spatiotemporal Hi-C data still have not been seen in the literature. We present HiC4D for dealing with the problem of forecasting spatiotemporal Hi-C data. We designed and benchmarked a novel network and named it residual ConvLSTM (ResConvLSTM), which is a combination of residual network and convolutional long short-term memory (ConvLSTM). We evaluated our new ResConvLSTM networks and compared them with the other five methods, including a naïve network (NaiveNet) that we designed as a baseline method and four outstanding video-prediction methods from the literature: ConvLSTM, spatiotemporal LSTM (ST-LSTM), self-attention LSTM (SA-LSTM) and simple video prediction (SimVP). We used eight different spatiotemporal Hi-C datasets for the blind test, including two from mouse embryogenesis, one from somatic cell nuclear transfer (SCNT) embryos, three embryogenesis datasets from different species and two non-embryogenesis datasets. Our evaluation results indicate that our ResConvLSTM networks almost always outperform the other methods on the eight blind-test datasets in terms of accurately predicting the Hi-C contact matrices at future time-steps. Our benchmarks also indicate that all of the methods that we benchmarked can successfully recover the boundaries of topologically associating domains called on the experimental Hi-C contact matrices. Taken together, our benchmarks suggest that HiC4D is an effective tool for predicting spatiotemporal Hi-C data. HiC4D is publicly available at both http://dna.cs.miami.edu/HiC4D/ and https://github.com/zwang-bioinformatics/HiC4D/.

Transcriptomic forecasting with neural ordinary differential equations

Single-cell transcriptomics technologies can uncover changes in the molecular states that underlie cellular phenotypes. However, understanding the dynamic cellular processes requires extending from inferring trajectories from snapshots of cellular states to estimating temporal changes in cellular gene expression. To address this challenge, we have developed a neural ordinary differential-equation-based method, RNAForecaster, for predicting gene expression states in single cells for multiple future time steps in an embedding-independent manner. We demonstrate that RNAForecaster can accurately predict future expression states in simulated single-cell transcriptomic data with cellular tracking over time. We then show that by using metabolic labeling single-cell RNA sequencing (scRNA-seq) data from constitutively dividing cells, RNAForecaster accurately recapitulates many of the expected changes in gene expression during progression through the cell cycle over a 3-day period. Thus, RNAForecaster enables short-term estimation of future expression states in biological systems from high-throughput datasets with temporal information.

A Neural Network Model to Estimate the Clogging of Ash Mixed Sand Filter

Infiltration filters of LID facilities are frequently confronted with clogging of pores by contaminants. To avoid malfunctioning of LID facilities, filters should be maintained according to the degree of clogging. In this study, clogging was characterized by curves representing the correlation between infiltration rate (&lt;i&gt;Q&lt;/i&gt;) and infiltration quantity (&lt;i&gt;IR&lt;/i&gt;). To estimate the degree of clogging, a regression neural network model was developed to estimate the label (infiltration quantity) corresponding to the selected features (ash mix ratio, infiltration rate, and variation of infiltration rate). According to the results obtained from the evaluation of accuracy and applicability of the final model trained over the whole training dataset, the following conclusions were drawn: 1) the training of the final model was verified to be correct according to mean absolute errors of 0.00445 and 0.00491 for the training and test datasets, respectively; 2) the final model could be applied to estimate multiple, multivariate clogging curves for arbitrary ash mixing ratios and trained range of time series in terms of infiltration rate; 3) given that the final model could not estimate the clogging curve for future time steps for which the model was not trained, a CNN model or LSTM-RNN model are suggested as an alternative.

A new method to identify non-steady thermal load based on element differential method

Inverse heat conduction problems (IHCPs) are widely used to identify the thermal load imposed on the components in aerospace fields, and the accuracy and robustness of identification algorithm are the keys to IHCP. In this paper, the element differential method (EDM), Levenberg-Marquardt method (LM) and future time steps method are combined to identify the thermal load. Firstly, the reliability and accuracy of the element differential method are fully verified in solving complex transient heat conduction problems of a variety of temperature-dependent materials. Then, the time-variant temperatures are obtained by the EDM with guessed thermal loads. And complex variable derivative method (CVDM) is used to compute the sensitivity matrix in LM. In addition, the influence of different initial values on the identified thermal loads of the considered 2D and 3D models is also fully analyzed. To prove the robustness of the proposed method, the noises are added to the measured data. Finally, the results show that the proposed method is effective in identifying complex temporal and spatial distributions of heat fluxes, and it shows good accuracy, robustness and noise resistance.

Fitting Imbalanced Uncertainties in Multi-output Time Series Forecasting

We focus on multi-step ahead time series forecasting with the multi-output strategy. From the perspective of multi-task learning (MTL), we recognize imbalanced uncertainties between prediction tasks of different future time steps. Unexpectedly, trained by the standard summed Mean Squared Error (MSE) loss, existing multi-output forecasting models may suffer from performance drops due to the inconsistency between the loss function and the imbalance structure. To address this problem, we reformulate each prediction task as a distinct Gaussian Mixture Model (GMM) and derive a multi-level Gaussian mixture loss function to better fit imbalanced uncertainties in multi-output time series forecasting. Instead of using the two-step Expectation-Maximization (EM) algorithm, we apply the self-attention mechanism on the task-specific parameters to learn the correlations between different prediction tasks and generate the weight distribution for each GMM component. In this way, our method jointly optimizes the parameters of the forecasting model and the mixture model simultaneously in an end-to-end fashion, avoiding the need of two-step optimization. Experiments on three real-world datasets demonstrate the effectiveness of our multi-level Gaussian mixture loss compared to models trained with the standard summed MSE loss function. All the experimental data and source code are available at https://github.com/smallGum/GMM-FNN .

Deep learning-based pedestrian trajectory prediction and risk assessment at signalized intersections using trajectory data captured through roadside LiDAR

In recent years, rapid advancements in the Autonomous Vehicles (AVs) industry have greatly motivated the research and development in pedestrian trajectory prediction and risk assessment. One of the critical requirements for AVs is to predict the future trajectories of pedestrians and provide collision warnings in an accurate and prompt manner. Nevertheless, accurate prediction of pedestrian trajectories remains a technical challenge, mainly caused by the heterogeneity of pedestrian crossing behavior and uncertainties in vehicle-pedestrian interactions. This paper proposes a deep learning-based method for pedestrian trajectory prediction and risk assessment, using trajectory data extracted from roadside LiDAR data and corresponding signal phasing information at MLK and Georgia Avenue in Chattanooga, TN. Meanwhile, a set of criteria referred to as the risk factor is established to quantitatively evaluate the risk of the pedestrian crossing behavior, which also serves as a learnable feature. A Long Short-Term Memory (LSTM) network is proposed, which takes the following data as the input: the pedestrian trajectory data, signal phasing data, and risk factors from the past 10 steps. Meanwhile, the network predicts the pedestrian trajectory and risk factor at the future time step. In the experimental study, the root-mean-square errors between the predicted and ground truth x and y coordinates are 0.225 meters and 0.377 meters, respectively, and the F1 score value for the risk factor is 99.6%, demonstrating the efficacy of the proposed LSTM-based methodology on pedestrian trajectory prediction and risk assessment.

Predictive Control Barrier Functions: Enhanced Safety Mechanisms for Learning-Based Control

While learning-based control techniques often outperform classical controller designs, safety requirements limit the acceptance of such methods in many applications. Recent developments address this issue through so-called predictive safety filters, which assess if a proposed learning-based control input can lead to constraint violations and modifies it if necessary to ensure safety for all future time steps. The theoretical guarantees of such predictive safety filters rely on the model assumptions and minor deviations can lead to failure of the filter putting the system at risk. This paper introduces an auxiliary soft-constrained predictive control problem that is always feasible at each time step and asymptotically stabilizes the feasible set of the original safety filter, thereby providing a recovery mechanism in safety-critical situations. This is achieved by a simple constraint tightening in combination with a terminal control barrier function. By extending discrete-time control barrier function theory, we establish that the proposed auxiliary problem provides a &#x2018;predictive&#x2019; control barrier function. The resulting algorithm is demonstrated using numerical examples.

Simultaneous estimation of time-varying boundary heat flux and temperature-dependent thermophysical and optical parameters of grey translucent medium

A hybrid optimization method, which combines unscented Kalman smoothing (UKS) technique and sequential quadratic programming (SQP) algorithm, is proposed to estimate the time-dependent boundary heat flux and temperature-dependent optical and thermophysical parameters (such as absorption coefficient, refractive index and thermal conductivity) of grey translucent medium simultaneously. The transient coupled radiative and conductive problem in grey translucent medium is solved by finite volume method (FVM) in forward model. In inverse problem, the measured signals consist of the temperature and radiative intensity on the boundaries of the medium. The effects of model parameters (such as the number of future time step and process noise covariance) and different incident heat fluxes on the reconstructed results are studied by the UKS-SQP technique. The results indicate that the proposed hybrid optimization technique is accurate and effective for determining the time-varying boundary heat flux, temperature-dependent optical and thermophysical parameters of grey translucent medium even with the measurement error.

Inverse Estimation of Transient Heat Flux Using Sequential Function Specification Method

Knowledge of surface heat flux is of great importance to numerous engineering fields. In this study, the sequential function specification method (SFSM) was employed to estimate the transient heat flux on the surface of semitransparent medium. The coupled radiation and conduction heat transfer was solved by discrete ordinate method combined with finite volume method. The temperature was used as input in estimation task which was solved by SFSM. Different types of heat fluxes were taken into account to test performance of the present technique. Retrieval results illustrated that the transient heat flux can be accurately determined using the SFSM. It was proved to be effective for different semitransparent media. For the heat flux with drastic changes, the SFSM showed some limitations. Moreover, the influences of future time step and measurement error on determination results were investigated. The SFSM was applicable even when the random deviation was increased to 1.5. The future time step of SFSM is suggested to be 3 for the estimation of transient heat flux on semitransparent medium surface.

Spatiotemporal Prediction Method for Vorticity of Three-dimensional Near-wall Cylindrical Flow Based on Convolutional Recurrent Neural Network

Due to the existence of a wall surface, the flow around the near-wall cylinder is a very complex unsteady dynamic behavior. However, this kind of feature extraction and fast prediction method of complex flow is difficult to solve in engineering. Therefore, a data-driven model based on Convolutional Recurrent Neural Network (CRNN) is proposed to solve the feature extraction and fast prediction of complex flow, and to predict the vorticity of the near wake region of a three-dimensional near-wall cylinder with G/D=1. CRNN model can analyze the relationship between Spatiotemporal evolution. The three-dimensional flow around the near-wall cylinder with Reynolds number 1500 meshed, and the vorticity was calculated after grid calculation to establish the Spatiotemporal evolution data set. For machine learning problems with high dimensions and large amounts of data, the Adaptive Moment (Adam) estimation method served as an effectively optimize. According to the Spatiotemporal prediction results of vorticity, the prediction results of the model fit well with the numerical simulation results, and the vorticity of the future time step is successfully predicted on the space and time series, that is, the evolutionary relationship of the Spatiotemporal series of vorticity near the wake of the near-wall cylinder is established. After the above model is established, only the initial multi-time vorticity data need to be obtained. The model can serve as predicting the vorticity field in the near wake region of the near-wall cylinder in the future.

Spatiotemporal Prediction of Radar Echoes Based on ConvLSTM and Multisource Data

Accurate and timely precipitation forecasts can help people and organizations make informed decisions, plan for potential weather-related disruptions, and protect lives and property. Instead of using physics-based numerical forecasts, which can be computationally prohibitive, there has been a growing interest in using deep learning techniques for precipitation prediction in recent years due to the success of these approaches in various other fields. These deep learning approaches generally use historical composite reflectivity (CR) at the surface level to predict future time steps. However, other relevant factors related to the potential motion and vertical structure of the storm have not been considered. To address this issue, this research proposes a multisource ConvLSTM (MS-ConvLSTM) model to improve the accuracy of precipitation forecasting by incorporating multiple data sources into the prediction process. The model was trained on a dataset of radar echo features, which includes not only composite reflectivity (CR), but also echo top (ET), vertically integrated liquid (VIL) water, and radar-retrieved wind field data at different elevations. Experiment results showed that the proposed model outperformed traditional methods in terms of various evaluation metrics, such as mean absolute error (MAE), mean squared error (MSE), probability of detection (POD), false alarm rate (FAR), and critical success index (CSI).

An unsupervised latent/output physics-informed convolutional-LSTM network for solving partial differential equations using peridynamic differential operator

This study presents a novel unsupervised convolutional Neural Network (NN) architecture with nonlocal interactions for solving Partial Differential Equations (PDEs). The nonlocal Peridynamic Differential Operator (PDDO) is employed as a convolutional filter for evaluating derivatives the field variable. The NN captures the time-dynamics in smaller latent space through encoder–decoder layers with a Convolutional Long–short Term Memory (ConvLSTM) layer between them. The ConvLSTM architecture is modified by employing a novel activation function to improve the predictive capability of the learning architecture for physics with periodic behavior. The physics is invoked in the form of governing equations at the output of the NN and in the latent (reduced) space. By considering a few benchmark PDEs, we demonstrate the training performance and extrapolation capability of this novel NN architecture by comparing against Physics Informed Neural Networks (PINN) type solvers. It is more capable of extrapolating the solution for future timesteps than the other existing architectures.

Inversion of spatio-temporal distribution heat flux and reconstruction of transient temperature field of three-layered skin tissue during hyperthermia

Hyperthermia (for example, high-intensity focused ultrasound, laser, radio-frequency) of cancerous cells from in vitro to in vivo requires accurately obtaining the heat distribution induced by external heating into the three-layered skin tissue. Obtaining the boundary heat flux into the three-layered skin tissue is a necessary condition to realize the measurement of tissue heat distribution. Considering the complexity of multiple boundary heat fluxes in spatio-temporal distribution, this study proposes an inversion scheme to predict the spatio-temporal distribution of multiple boundary heat fluxes into the three-layered skin tissue. In the inversion scheme, a multivariable prediction model is established to solve the spatio-temporal coupling problem between the inversed boundary heat flux and measurement temperature information. Furthermore, based on the dependence between the predicted temperature and inversed boundary heat flux, the inversion system is constructed to realize the simultaneous optimization inversion of multiple boundary heat fluxes in spatio-temporal distribution. To examine the feasibility and effectiveness of inversion scheme, numerical experiments are carried out to discuss the influence of future time steps and measurement errors on the inversion results of boundary heat flux. In addition, the transient temperature field of three-layered skin tissue is reconstructed by inversed boundary heat flux, which could provide an economical, effective, and non-invasive solution for the measurement of thermal field of three-layered skin tissue during hyperthermia.

Deep Learning for Wind Speed Forecasting Using Bi-LSTM with Selected Features

Wind speed forecasting is important for wind energy forecasting. In the modern era, the increase in energy demand can be managed effectively by forecasting the wind speed accurately. The main objective of this research is to improve the performance of wind speed forecasting by handling uncertainty, the curse of dimensionality, overfitting and non-linearity issues. The curse of dimensionality and overfitting issues are handled by using Boruta feature selection. The uncertainty and the non-linearity issues are addressed by using the deep learning based Bi-directional Long Short Term Memory (Bi-LSTM). In this paper, Bi-LSTM with Boruta feature selection named BFS-Bi-LSTM is proposed to improve the performance of wind speed forecasting. The model identifies relevant features for wind speed forecasting from the meteorological features using Boruta wrapper feature selection (BFS). Followed by Bi-LSTM predicts the wind speed by considering the wind speed from the past and future time steps. The proposed BFS-Bi-LSTM model is compared against Multilayer perceptron (MLP), MLP with Boruta (BFS-MLP), Long Short Term Memory (LSTM), LSTM with Boruta (BFS-LSTM) and Bi-LSTM in terms of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Square Error (MSE) and R2. The BFS-Bi-LSTM surpassed other models by producing RMSE of 0.784, MAE of 0.530, MSE of 0.615 and R2 of 0.8766. The experimental result shows that the BFS-Bi-LSTM produced better forecasting results compared to others.

Temporal predictions of periodic flows using a mesh transformation and deep learning-based strategy

This paper focuses on the temporal prediction of unsteady flow based on a combination of mesh transformation and deep learning technology. To this end, a body-fitted mesh interpolation was first implemented to extract the flow information in the region of interest and maintain the detailed flow information in the boundary layer. Subsequently, a mesh transformation technique was applied to map the physical coordinates into the curvilinear coordinates to obtain a structural uniform mesh, which is facilitated for the implementation the structural neural networks. Ultimately, two advanced neural networks, i.e., Unet and Fourier Neural Operator (FNO), were used for establishing the relationship between flow at previous time steps and that at future time steps. Two flow configurations numerically simulated—i.e., a laminar flow around a cylinder at Re=200 and a transonic buffet of a supercritical airfoil at Ma=0.73—were selected as the benchmarks to demonstrate the effectiveness and robustness of the proposed strategy. The results indicated that the flow at future time steps can be accurately obtained using that at the previous time steps via both neural networks, even in a complicated flow scenario. The velocity and pressure of the periodic flow can be faithfully predicted with a remarkable accuracy. Additionally, the derived flow physical quantities were comprehensively compared, which demonstrated that better performance can be achieved using FNO with a lower relative loss and a shorter training time-consuming.

Leveraging Graph and Deep Learning Uncertainties to Detect Anomalous Maritime Trajectories

Understanding and representing traffic patterns are key to detecting anomalous trajectories in the transportation domain. However, some trajectories can exhibit heterogeneous maneuvering characteristics despite confining to normal patterns. Thus, we propose a novel graph-based trajectory representation and association scheme for extraction and confederation of traffic movement patterns, such that data patterns and uncertainty can be learned by deep learning (DL) models. This paper proposes the usage of a recurrent neural network (RNN)-based evidential regression model, which can predict trajectory at future timesteps as well as estimate the data and model uncertainties associated, to detect anomalous maritime trajectories, such as unusual vessel maneuvering, using automatic identification system (AIS) data. Furthermore, we utilize evidential deep learning classifiers to detect unusual turns of vessels and the loss of transmitted signal using predicted class probabilities with associated uncertainties. Our experimental results suggest that the graphical representation of traffic patterns improves the ability of the DL models, such as evidential and Monte Carlo dropout, to learn the temporal-spatial correlation of data and associated uncertainties. Using different datasets and experiments, we demonstrate that the estimated prediction uncertainty yields fundamental information for the detection of traffic anomalies in the maritime and, possibly in other domains.

MTL-Deep-STF: A multitask learning based deep spatiotemporal fusion model for outdoor air temperature prediction in building HVAC systems

Buildings consume large quantities of energy. Reducing building energy consumption is essential to achieving carbon neutrality goals. Building energy consumption is strongly influenced by various meteorological factors for which outdoor air temperature is a proxy. This study proposes a data-driven multitask-learning deep spatiotemporal fusion model (MTL-Deep-STF) that predicts outdoor air temperature to provide baseline data against which improvements to conserve energy in heating, ventilation, and air conditioning (HVAC) systems can be proposed and assessed. MTL-Deep-STF is a generic multiple input–multiple output model that predicts temperatures for successive time points in separate high-level tasks. Each task uses a multilayer perceptron to combine meteorological data from neighboring observation stations. CNN-GRU is used to extract spatially local long-term and short-term temporal features that are shared between different tasks via CNN. Comparative validation experiments using real world datasets for complex terrains showed that the model was highly accurate in both single step and multistep predictions. One-step-ahead predictions for winter (MAE = 0.428 °C, RMSE = 0.705, Acc ≤2 °C = 97.832%) and summer (MAE = 0.467 °C, RMSE = 0.706, Acc ≤2 °C = 97.899%) datasets were particularly accurate. This accuracy was coupled with high robustness ( R 2 = 0.976 for winter and R 2 = 0.981 for summer). Use of the model can conserve building energy. • The spatiotemporal fusion model MTL-Deep-STF uses data from neighboring stations to predict air temperature at multiple future time steps. • MTL-Deep-STF treats temperature prediction for successive prediction horizons as distinct tasks. • MTL-Deep-STF extracts global spatial features of multiple stations through fully connected networks. • MTL-Deep-STF is extremely accurate and robust in predicting outdoor temperatures and is therefore suitable for controlling HVAC systems, particularly in complex terrains.

Supervised ML Solution for Band Assignment in Dual-Band Systems With Omnidirectional and Directional Antennas

Many wireless networks, including 5G NR (New Radio) and future beyond 5G cellular systems, are expected to operate on multiple frequency bands. This paper considers the band assignment (BA) problem in dual-band systems, where the basestation (BS) chooses one of the two available frequency bands (centimeter-wave and millimeter-wave bands) to communicate with the user equipment (UE). While the millimeter-wave band might offer higher data rate, there is a significant probability of outage during which the communication should be carried on the (more reliable) centimeter-wave band. With mobility, the BA can be perceived as a sequential problem, where the BS uses previously observed information to predict the best band for a future time step. We formulate the BA as a binary classification problem and propose supervised Machine Learning (ML) solutions. We study the problem when both the BS and the UE use (i) omnidirectional antennas and (ii) both use directional antennas. In the omnidirectional case, we derive analytical benchmark solutions based on the Gaussian Process (GP) assumption for the inter-band shadow fading. In the directional case, where the labeling is shown to be complex, we propose an efficient labeling approach based on the Viterbi Algorithm (VA). We compare the performances for two channel models: (i) a stochastic channel and (ii) a ray-tracing based channel.