Prediction Of Timing Research Articles

Optimizing throughput of Seq2Seq model training on the IPU platform for AI-accelerated CFD simulations

Intelligence Processing Units (IPU) have proven useful for many AI applications. In this paper, we evaluate them within the emerging field of AI for simulation, where traditional numerical simulations are supported by artificial intelligence approaches. We focus specifically on a program for training machine learning models supporting a computational fluid dynamics application. We use custom TensorFlow provided by the Poplar Software Development Kit to adapt the program for the IPU-POD16 platform and investigate its ease of use and performance scalability. Training a model on data from OpenFOAM simulations allows us to get accurate simulation state predictions in test time. We describe how to optimize multi-threading runtime options and utilize the popdist library to overcome a performance bottleneck in feeding training data to the IPU on the host side. Due to communication overheads, using data parallelism to utilize two IPUs instead of one does not improve the throughput. However, once the intra-IPU costs have been paid, the hardware capabilities for inter-IPU communication allow for good scalability. Increasing the number of IPUs from two to 16 improves the throughput from 560.8 to 2805.8 samples/s. Additionally, the experimental results show that reducing the precision of input data storage from FP32 to FP16 allows to improve training throughput by 12%, while tuning selected runtime variables, by up to 6.3%.

An implementation of XGBoost algorithm to estimate effective porosity on well log data

The application of machine learning methods is aimed at providing efficiency and avoiding subjectivity in estimating reservoir porosity data. This study proposes the eXtreme Gradient Boost (XGBoost) algorithm which is known to be effective in providing accurate predictions in a short time for estimating effective porosity. The model was optimized using the GridSearchCV (GS) module, then applied to 7 wells from Damar field, Indonesia with variations in the separation of training and testing data based on the number of wells. The best evaluation results are achieved when uses 6 training wells and one tested well with a model accuracy around 78.36% and training time around 1.29 seconds. An increasing the amount of training data will increase the model performance. All variations of training and testing did not show any indication of overfitting. Therefore, it can be concluded that the XGBoost model can effectively estimate reservoir porosity in the area.

Music Recommendation System Using Deep Learning

Abstract: As part of this study, we present a music recommendation system that makes use of deep learning. The system learns a neural network that can anticipate the user's musical preferences and provide personalised playlist suggestions based on the user's listening habits. The proposed method considers both the user's explicit tastes and those that may be deduced from their listening patterns. The technology is able to adapt to a user's evolving preferences over time and provide more relevant recommendations. Software that predicts what a user will wish to buy based on their tastes and prior purchases is called a recommendation system. Although the emphasis of this paper is on improving music recommendation systems, the approach outlined here might be applied to a broad range of other platforms and domains as well, including video sharing sites like YouTube and Netflix as well as online retailers like Amazon. System efficiency decreases as complexity increases. Our technique, Tunes Recommendation System (T-RECSYS), provides an efficient recommendation system that can make predictions in real time by integrating data from both content-based and collaborative filtering into a deep learning classification model. By applying our strategy to the Spotify Recsys Challenge dataset, we find a threshold that provides an optimal trade-off between false positives and false negatives, increasing accuracy to 88%

Clinical use of the canine foetal kidney formula in dogs of different maternal sizes during the last ten days before parturition.

Data concerning the use of the canine foetal kidney length (L) formula in the prediction of parturition timing are still lacking. In our study, we aimed to evaluate the accuracy of the L formula in predicting the parturition date during the last 10 days of pregnancy. Twenty-five clinically healthy pregnant bitches, aged 2-9 years and weighing 3.5-52.2kg, were ultrasonographically monitored from -11 to 0 days before parturition (dbp). Kidney L was measured for the three most caudal foetuses, and the parturition day was estimated using the kidney formula, whose accuracy was calculated as the percentage of cases estimated (ranges of ± 1 or ± 2 days) on the actual parturition date. A K-proportions test was performed to identify differences in the accuracy among maternal sizes and the sex ratio of pups, and a two-proportions z-test was performed to identify differences between litter size classes (≤ 7 vs. > 7 pups) and time ranges (-11/-5 and -4/0 dbp). An accuracy of 35% within ± 2 days was found in the range of -11/-5 dbp and an accuracy of 30% within ± 2 days was found in the range of -4/0 dbp. The accuracy differed between small (53% ±1day and 60% ±2 days) and large (10% within ± 1 and ± 2 days) bitches (P = 0.019 within ± 1day, and P = 0.007 within ± 2 days). For small litter sizes, the accuracy was 38% within ± 1day and 44% within ± 2 days, and for large litter sizes, it was 14% within ± 1 and ± 2 days. A threshold value was found between litter size classeswithin ± 2 days. The use of the L formula during the last ten days of pregnancy did not seem to warrant good accuracy in the prediction of parturition date. Further studies on different maternal sizes should be performed.

Evaluation of the wellbore drillability while horizontally drilling sandstone formations using combined regression analysis and machine learning models

The rate of penetration (ROP) is an influential parameter in the optimization of oil well drilling because it has a huge impact on the total drilling cost. This study aims to optimize four machine learning models for real-time evaluation of the ROP based on drilling parameters during horizontal drilling of sandstone formations. Two well data sets were implemented for the model training–testing (Well-X) and validation (Well-Y). A total of 1224 and 524 datasets were implemented for training and testing the model, respectively. A correlation for ROP assessment was suggested based on the optimized artificial neural network (ANN) model. The precision of this equation and the optimized models were tested (524 datapoints) and validated (2213 datapoints), and their accuracy was compared to available ROP correlations. The developed ANN-based equation predicted the ROP with average absolute percentage errors (AAPE) of 0.3% and 1.0% for the testing and validation data, respectively. The new empirical equation and the optimized fuzzy logic and functional neural network models outperformed the available correlations in assessing the ROP. The support vector regression accuracy performance showed AAPE of 26.5%, and the correlation coefficient for the estimated ROP was 0.50 for the validation phase. The outcomes of this work could help in modeling the ROP prediction in real time during the drilling process.

Minimising Butadiene Level in Liquefied Petroleum Gas (LPG) via Non-Stirred Blending with Numerical Approach

LPG has variable commercial grades due to its varying LPG composition will yield different properties. Because their properties depend on the composition, the LPG quality will differ based on the source of its production extracted either from refinery streams or natural gas. The composition of propane and butane, which are the two major components in LPG composition, play an important role in reflecting the properties of LPG, as the mixtures influence the boiling points and also meet product specification requirements. Butadiene is also present in LPG, albeit in trace amounts. Although butadiene is a minor component, it must be kept minimised of the total weight fraction due to regulatory limits. Butadiene is a hazardous chemical that, when inhaled, can cause cancer and genetic defects. LPG used for commercial purposes that contain of any carcinogenic substance such as butadiene must also be classified as carcinogenic. The LPG plant operator is facing the problem that imported LPG composition from outside sources contain levels of butadiene that exceed the regulatory limit of 0.5% of weight fraction. LPG composition, containing butadiene levels that exceed 0.5% of weight fraction is considered as off-specification, while butadiene levels less than 0.5% of weight fraction are considered as on-specification the LPG products. To reduce the levels of butadiene that exceed 0.5% of the weight fraction in off-specification LPG products, the blending of on-specification LPG products with off-specification was introduced and provided the most economical method inside the plant. The jet mixing approach was selected to predict the homogeneous blending time for each different off-specification LPG composition because it is the best approximation for natural mixing behaviour. Four empirical mixing time correlations of jet mixing were applied for the prediction of homogeneous blending time; the correlations were derived by Lane and Rice (1982), Maruyama, Ban and Mizushina (1982), Grenville and Tilton (1997) and Hiby and Modigell (1978). The homogeneous blending time predicted by these four mixing time correlations decreased as the quantity of on-specification LPG required increased, which is in good agreement with both simulation results of 95% and 99% mixing. Therefore, due to the development of individual jet mixer correlation, these four mixing time correlation results in different homogeneous blending times was going through the different measurement techniques and monitoring methods in jet mixed tanks.

EFFNet-CA: An Efficient Driver Distraction Detection Based on Multiscale Features Extractions and Channel Attention Mechanism.

Driver distraction is considered a main cause of road accidents, every year, thousands of people obtain serious injuries, and most of them lose their lives. In addition, a continuous increase can be found in road accidents due to driver's distractions, such as talking, drinking, and using electronic devices, among others. Similarly, several researchers have developed different traditional deep learning techniques for the efficient detection of driver activity. However, the current studies need further improvement due to the higher number of false predictions in real time. To cope with these issues, it is significant to develop an effective technique which detects driver's behavior in real time to prevent human lives and their property from being damaged. In this work, we develop a convolutional neural network (CNN)-based technique with the integration of a channel attention (CA) mechanism for efficient and effective detection of driver behavior. Moreover, we compared the proposed model with solo and integration flavors of various backbone models and CA such as VGG16, VGG16+CA, ResNet50, ResNet50+CA, Xception, Xception+CA, InceptionV3, InceptionV3+CA, and EfficientNetB0. Additionally, the proposed model obtained optimal performance in terms of evaluation metrics, for instance, accuracy, precision, recall, and F1-score using two well-known datasets such as AUC Distracted Driver (AUCD2) and State Farm Distracted Driver Detection (SFD3). The proposed model achieved 99.58% result in terms of accuracy using SFD3 while 98.97% accuracy on AUCD2 datasets.

Dynamic effects of prognostic factors and individual survival prediction for amyotrophic lateral sclerosis disease.

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease affecting motor neurons, with broad heterogeneity in disease progression and survival in different patients. Therefore, an accurate prediction model will be crucial to implement timely interventions and prolong patient survival time. A total of 1260 ALS patients from the PRO-ACT database were included in the analysis. Their demographics, clinical variables, and death reports were included. We constructed an ALS dynamic Cox model through the landmarking approach. The predictive performance of the model at different landmark time points was evaluated by calculating the area under the curve (AUC) and Brier score. Three baseline covariates and seven time-dependent covariates were selected to construct the ALS dynamic Cox model. For better prognostic analysis, this model identified dynamic effects of treatment, albumin, creatinine, calcium, hematocrit, and hemoglobin. Its prediction performance (at all landmark time points, AUC ≥ 0.70 and Brier score ≤ 0.12) was better than that of the traditional Cox model, and it predicted the dynamic 6-month survival probability according to the longitudinal information of individual patients. We developed an ALS dynamic Cox model with ALS longitudinal clinical trial datasets as the inputs. This model can not only capture the dynamic prognostic effect of both baseline and longitudinal covariates but also make individual survival predictions in real time, which are valuable for improving the prognosis of ALS patients and providing a reference for clinicians to make clinical decisions.

Modality-Attention Promotes the Neural Effects of Precise Timing Prediction in Early Sensory Processing.

Precise timing prediction (TP) enables the brain to accurately predict the occurrence of upcoming events in millisecond timescale, which is fundamental for adaptive behaviors. The neural effect of the TP within a single sensory modality has been widely studied. However, less is known about how precise TP works when the brain is concurrently faced with multimodality sensory inputs. Modality attention (MA) is a crucial cognitive function for dealing with the overwhelming information induced by multimodality sensory inputs. Therefore, it is necessary to investigate whether and how the MA influences the neural effects of the precise TP. This study designed a visual-auditory temporal discrimination task, in which the MA was allocated to visual or auditory modality, and the TP was manipulated into no timing prediction (NTP), matched timing prediction (MTP), and violated timing prediction (VTP) conditions. Behavioral and electroencephalogram (EEG) data were recorded from 27 subjects, event-related potentials (ERP), time-frequency distributions of inter-trial coherence (ITC), and event-related spectral perturbation (ERSP) were analyzed. In the visual modality, precise TP led to N1 amplitude and 200-400 ms theta ITC variations. Such variations only emerged when the MA was attended. In auditory modality, the MTP had the largest P2 amplitude and delta ITC than other TP conditions when the MA was attended, whereas the distinctions disappeared when the MA was unattended. The results suggest that the MA promoted the neural effects of the precise TP in early sensory processing, which provides more neural evidence for better understanding the interactions between the TP and MA.

Genetic and Microenvironment Features Do Not Distinguish Follicular Lymphoma Patients Requiring Immediate or Deferred Treatment.

Genetic and Microenvironment Features Do Not Distinguish Follicular Lymphoma Patients Requiring Immediate or Deferred Treatment.

CGE23-070: Baseline Neutrophil to Lymphocyte Ratio and Its Change Over Time Predict Overall Survival in Metastatic Non–Small Cell Lung Cancer Patients Who Received First Line Pembrolizumab Therapy

"CGE23-070: Baseline Neutrophil to Lymphocyte Ratio and Its Change Over Time Predict Overall Survival in Metastatic Non–Small Cell Lung Cancer Patients Who Received First Line Pembrolizumab Therapy" published on 31 Mar 2023 by National Comprehensive Cancer Network.

Continual spatio-temporal graph convolutional networks

Graph-based reasoning over skeleton data has emerged as a promising approach for human action recognition. However, the application of prior graph-based methods, which predominantly employ whole temporal sequences as their input, to the setting of online inference entails considerable computational redundancy. In this paper, we tackle this issue by reformulating the Spatio-Temporal Graph Convolutional Neural Network as a Continual Inference Network, which can perform step-by-step predictions in time without repeat frame processing. To evaluate our method, we create a continual version of ST-GCN, CoST-GCN, alongside two derived methods with different self-attention mechanisms, CoAGCN and CoS-TR. We investigate weight transfer strategies and architectural modifications for inference acceleration, and perform experiments on the NTU RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400 datasets. Retaining similar predictive accuracy, we observe up to 109× reduction in time complexity, on-hardware accelerations of 26×, and reductions in maximum allocated memory of 52% during online inference.

Predictive mechanisms in adaptive chunking during rapid extraction of meaning from spoken signals

Speech understanding requires rapid of categorizations of auditory inputs. Neuronal rhythms play key roles in these computations; however, the principles by which timing, acoustics, and context are combined on-line to accomplish speech perception remain murky. Here, we outline a model of how temporal constraints sculpt meaning-making in speech processing according to an adaptively quasi-rhythmic process driven by fluctuations in certainty and predictability of timing and content. We propose that predictive mechanisms are leveraged to reduce possible identities of inputs based on prior inputs in which the smallest chunks of input can be rapidly compressed, recoded, and passed to higher representational linguistic levels. Evidence accumulation at each level of representation is used to assess likelihoods of candidate interpretations and their reliabilities at different timescales in a manner pegged to speech rate and situation-specific speed-accuracy tradeoffs. This synthesis illuminates mechanisms of human speech processing while making predictions for neuronal implementations and behavioral psychophysics.

High Resolution Forecasting of Summer Drought in the Western United States

AbstractDrought monitoring and forecasting systems are used in the United States (U.S.) to inform drought management decisions. Drought forecasting efforts have often been conducted and evaluated at coarse spatial resolutions (i.e., >10‐km), which miss key local drought information at higher resolutions. Addressing the importance of forecasting drought at high resolutions, this study develops statistical models to evaluate 1‐ to 3‐month lead time predictability of meteorological and agricultural summer drought across the western U.S. at a 4‐km resolution. Our high‐resolution drought predictions have statistically significant skill (p ≤ 0.05) across 70%–100% of the western U.S., varying by evaluation metric and lead time. 1‐ to 3‐month lead time drought forecasts accurately represent monitored summer drought spatial patterns during major drought events, the interannual variability of drought area from 1982 to 2020 (r = 0.84–0.93), and drought trends (r = 0.94–0.97). 71% of western U.S summer drought area interannual variability can be explained by cold‐season (November–February) climate conditions alone allowing skillful 3‐month lead time predictions. Pre‐summer drought conditions (represented by drought indices) are the most important predictors for summer drought. Thus, the statistical models developed in this study heavily rely on the autocorrelation of chosen agricultural and meteorological drought indices which estimate land surface moisture memory. Indeed, prediction skill strongly correlates with persistence of drought conditions (r ≥ 0.73). This study is intended to support future development of operational drought early warning systems that inform drought management.

Natural cycle frozen embryo transfer: a survey of current assisted reproductive technology practices in the U.S.

Emerging data suggests improved obstetric outcomes with frozen embryo transfer (FET) in an ovulatory or natural cycle (NC-FETs), as compared to programmed endometrial preparation. The objective of this study is to better understand practice patterns and provider attitudes regarding the use of NC-FETs in the United States (U.S.). In this cross-sectional study, an anonymous 22-question survey was emailed to 441 U.S. Assisted Reproductive Technology (ART) clinics to assess the utilization of NC endometrial preparation for FET, protocols used, restrictions to offering NC-FET, and providers' perspectives on advantages and disadvantages of NC-FET. Descriptive statistics were used to analyze survey responses. The survey response rate was 49% (216/441). Seventeen percent of responding clinics did not offer NC-FET. Of the clinics that did not offer NC-FET, 65% had only 1-2 physicians in their practice. Common reasons for not offering NC-FET included "lack of timing predictability for transfer" (81%) and "increased burden on staff/laboratory personnel on holidays and weekends" (54%). Of clinics offering NC-FET, 76% reported < 25% of cycles used the NC for endometrial preparation. Over half (52%) of clinics that offered NC-FET reported having eligibility restrictions for NC-FET. Reported benefits of NC-FET were "patient satisfaction" (18%), "decreased cost of medications" (18%), and "avoidance of intramuscular progesterone" (17%). The attitude towards NC-FET in their clinics was reported as positive by 65% of respondents. NC-FETs are offered by most U.S. ART clinics but are used only in the minority of FET cycles for endometrial preparation, and use is often restricted.

Early anomaly detection in time series: a hierarchical approach for predicting critical health episodes

The early detection of anomalous events in time series data is essential in many domains of application. In this paper we deal with critical health events, which represent a significant cause of mortality in intensive care units of hospitals. The timely prediction of these events is crucial for mitigating their consequences and improving healthcare. One of the most common approaches to tackle early anomaly detection problems is through standard classification methods. In this paper we propose a novel method that uses a layered learning architecture to address these tasks. One key contribution of our work is the idea of pre-conditional events, which denote arbitrary but computable relaxed versions of the event of interest. We leverage this idea to break the original problem into two hierarchical layers, which we hypothesize are easier to solve. The results suggest that the proposed approach leads to a better performance relative to state of the art approaches for critical health episode prediction.

Numerical investigation of a large bore, direct injection, spark ignition, hydrogen-fuelled engine

This paper presents Unsteady Reynolds Averaged Navier Stokes (URANS) simulations of a large bore, hydrogen-fuelled direct injection spark ignition (DISI) engine at different spark and start of injection (SOI) timings. Six cases are simulated, including three with various spark timings at a low boost level and three with advanced to late injection timings at a higher boost level. The numerical simulations are validated with experimental data for four out of six cases, while the other two are considered blind computational fluid dynamics (CFD) simulations. It is shown that the autoignition occurs with advanced spark timing due to high in-cylinder pressure and unburnt temperature. For different SOIs, it is demonstrated that flame propagation involves a spark-initiated flame combined with an autoignition generated flame. The case with the late injection timing features poor mixing and slower combustion due to the presence of lean mixtures near the spark plug. As a result, this case features the lowest thermal efficiency when SOI is varied. In all cases, both mixture and temperature stratification are present. Simulations of zero-dimensional chemical reactors demonstrate that this stratification must be correctly captured for accurate prediction of autoignition timing.

Flooding time nomograph for urban river flood prediction: Case study of Dorim stream basin, Seoul

AbstractGlobal climate change is intensifying flood damage in urban rivers. Notably, most small and medium‐sized urban rivers have a brief concentration period and are highly vulnerable to sudden heavy rains that lead to a rapid increase in water levels. Therefore, rapid flood forecasting must be performed through accurate flood occurrence and timing prediction. In this study, a flooding time nomograph (FTN) was proposed to predict the flood occurrence time according to rainfall conditions, such as intensity, time distribution, and duration. In addition, rainfall–runoff simulations were performed by establishing different virtual rainfall scenarios using Huff's quartile rainfall time distribution. The simulation results were used to formulate the relationship between the rainfall intensity and flood occurrence time to generate the FTN. The applicability of this tool was verified through a comparison with the observed flood occurrence time for an actual rainfall event, which was highly accurate in the target watershed, with a correlation coefficient &gt;0.8 and Nash–Sutcliffe efficiency &gt;0.6. Therefore, the proposed FTN can be used to reasonably predict the occurrence of floods and the time of flood occurrence using only predicted rainfall information.

A Function Approximator Model for Robust Online Foot Angle Trajectory Prediction Using a Single IMU Sensor: Implication for Controlling Active Prosthetic Feet

Lower limb dysfunction hinders the locomotive activity in individuals and compels them to use assistive devices such as prostheses or orthoses to restore the missing locomotive functions. Passive assistive devices have limitations in replacing the lost functionality. On the other hand, active devices could restore more natural locomotion using motorized joints. To make proper use of these embedded motors, a control architecture is required to translate the motion intent of the user into the intended prosthesis joint trajectories. In this article, we propose a temporal convolution-based online foot angle trajectory prediction network (FATP-N) that predicts the foot angular positions during unconstrained walking from the shank angular position measured using a single wearable motion tracker sensor. The data acquisition experiments were designed to reflect natural gait with turns, varying inclines, speeds, and mixed cadences. The trained models were prepared for real-time prediction by compressing them to reduce the storage and time complexity. The validations were performed on different motion conditions, including those that have not been used for model training, to verify the robustness of the model. The proposed FATP-N achieved high accuracy and could generate predictions in real time. The results indicate that the proposed approach offers a possible architecture for real-time control of powered intelligent prostheses.

Slow Slip Detectability in Seafloor Pressure Records Offshore Alaska

AbstractIn subduction zones worldwide, seafloor pressure data are used to observe tectonic deformation, particularly from megathrust earthquakes and slow slip events (SSEs). However, such measurements are also sensitive to oceanographic circulation‐generated pressures over a range of frequencies that conflate with tectonic signals of interest. Using seafloor pressure and temperature data from the Alaska Amphibious Community Seismic Experiment, and sea surface height data from satellite altimetry, we evaluate the efficacy of various seasonal and oceanographic pressure signal proxy corrections and conduct synthetic tests to determine their impact on the timing and amplitude prediction of ramp‐like signals typical of SSEs. We find that subtracting out the first mode of the complex empirical orthogonal functions of the pressure records on either the shelf or slope yields signal root‐mean‐square error (RMS) reductions up to 73% or 80%, respectively. Additional correction with proxies that exploit the depth‐dependent spatial coherence of pressure records provides cumulative variance reductions up to 83% and 93%, respectively. Our detectability tests show that the timing and amplitude of synthetic SSE‐like ramps can be well constrained for ramp amplitudes ≥4 cm on the shelf and ≥2 cm on the slope, using a fully automated detector. The principal limits on detectability are residual abrupt changes in pressure that occur as part of the transition to and from summer to winter conditions but are not adequately characterized by our seasonal corrections, as well as the inability to properly account for instrumental drift, which is not readily separated from the seasonal signal.