Time-based Model Research Articles

Simplifying Distributed Neural Network Training on Massive Graphs: Randomized Partitions Improve Model Aggregation

Distributed Graph Neural Network (GNN) training facilitates learning on massive graphs that surpass the storage and computational capabilities of a single machine. Traditional distributed frameworks strive for performance parity with centralized training by maximally recovering cross-instance node dependencies, relying either on inter-instance communication or periodic fallback to centralized training. However, these processes create overhead and constrain the scalability of the framework. In this work, we propose a streamlined framework for distributed GNN training that eliminates these costly operations, yielding improved scalability, convergence speed, and performance over state-of-the-art approaches. Our framework (1) comprises independent trainers that asynchronously learn local models from locally-available parts of the training graph, and (2) synchronizes these local models only through periodic (time-based) model aggregation. Contrary to prevailing belief, our theoretical analysis shows that it is not essential to maximize the recovery of cross-instance node dependencies to achieve performance parity with centralized training. Instead, our framework leverages randomized assignment of nodes or super-nodes (i.e., collections of original nodes) to partition the training graph in order to enhance data uniformity and minimize discrepancies in gradient and loss function across instances. Experiments on social and e-commerce networks with up to 1.3 billion edges show that our proposed framework achieves state-of-the-art performance and 2.31x speedup compared to the fastest baseline, despite using less training data.

On the Model Update Strategies for Supervised Learning in AIOps Solutions

AIOps (Artificial Intelligence for IT Operations) solutions leverage the massive data produced during the operation of large-scale systems and machine learning models to assist software engineers in their system operations. As operation data produced in the field are constantly evolving due to factors such as the changing operational environment and user base, the models in AIOps solutions need to be constantly maintained after deployment. While prior works focus on innovative modeling techniques to improve the performance of AIOps models before releasing them into the field, when and how to update AIOps models remain an under-investigated topic. In this work, we performed a case study on three large-scale public operation data: two trace datasets from the cloud computing platforms of Google and Alibaba and one disk stats dataset from the BackBlaze cloud storage data center. We empirically assessed five different types of model update strategies for supervised learning regarding their performance, updating cost, and stability. We observed that active model update strategies (e.g., periodical retraining, concept drift guided retraining, time-based model ensembles, and online learning) achieve better and more stable performance than a stationary model. Particularly, applying sophisticated model update strategies (e.g., concept drift detection, time-based ensembles, and online learning) could provide better performance, efficiency, and stability than simply retraining AIOps models periodically. In addition, we observed that, although some update strategies (e.g., time-based ensemble and online learning) can save model training time, they significantly sacrifice model testing time, which could hinder their applications in AIOps solutions where the operation data arrive at high pace and volume and where immediate inferences are required. Our findings highlight that practitioners should consider the evolution of operation data and actively maintain AIOps models over time. Our observations can also guide researchers and practitioners in investigating more efficient and effective model update strategies that fit in the context of AIOps.

Identifying immune-related prognostic biomarkers in osteosarcoma: Development and validation of the tumor immune microenvironment risk model based on analysis of TCGA.

Osteosarcoma is a rare and aggressive malignancy with limited effective therapeutic options. This study aimed to identify immune-related prognostic biomarkers and develop a prognostic model for osteosarcoma. We performed integrated analysis of transcriptomic data and immune cell infiltration profiles of 84 osteosarcoma samples from the Cancer Genome Atlas (TCGA) database. Time-dependent receiver operating characteristic (ROC) curve analysis was used to assess the prognostic value of the TIMErisk model. We also performed functional annotation and pathway enrichment analyses to explore the potential mechanisms underlying the TIMErisk model. We identified a seven-gene TIMErisk model (C2, APBB1IP, BST2, TRPV2, CCL5, GBP1, and F13A1) that was independently associated with overall survival of osteosarcoma patients. The TIMErisk model showed significant associations with immune cell infiltration and immunosuppressive gene expression. In addition, the TIMErisk model was associated with drug sensitivity, and we found that several immune checkpoint genes were significantly differentially expressed between high- and low-TIMErisk groups. Functional annotation and pathway enrichment analyses revealed that the TIMErisk model was associated with multiple immune-related pathways, including antigen processing and presentation, cytokine-cytokine receptor interaction, and T cell receptor signaling pathway. Our study identified a novel TIME-based prognostic model for osteosarcoma that incorporates immune-related genes and can be used to predict patient prognosis and response to immunotherapy. Our findings highlight the importance of the TIME microenvironment in osteosarcoma progression and suggest that immune-related biomarkers may have clinical significance in the management of osteosarcoma.

Residual life modeling and maintenance planning for repairable systems

Abstract Accepted by: Phil Scarf This paper formulates a state-dependent mean residual lifetime model for a repairable system operating in a dynamic environment. The problem is addressed by means of a two-state damage process reflecting the effect of operating environment on the system and a repair process associated with the damage process. As the damage process shifts to a higher state, to maintain a minimum level of performance, the decision maker repairs the system at times that arise according to a point process with a constant intensity. We demonstrate the generality of the proposed model and show how existing models emerge as specific cases. Our approach stimulates further research on the determination of two types of maintenance policies: maintenance policy based on the number of imperfect repairs (Model I) and time-based maintenance policy (Model II). In both cases, using the renewal reward theorem argument, we aim at minimizing the long-run average maintenance cost per unit time by determining optimal replacement policies and the optimal level of imperfect repairs. We illustrate the proposed models and carry out a comparative analysis of maintenance policies through numerical examples. The main conclusions drawn are that repair and maintenance policies depend on the failure mechanism, repair frequency and the level of costs involved. Also, numerical comparison shows that the maintenance modelling based on the number of imperfect repairs (model I) outperforms the time-based replacement model (model II) and two baseline models ignoring the effect of operating environment or whose attention is restricted to perfect repair.

Time Out: Auditor Perceptions and Audit Outcomes When Moving Away from a Time-Based Budget Audit Model

SYNOPSIS The purpose of this study is to examine the benefits and costs of an alternative to the traditional time-based budget audit model. We survey auditors who participated in a real-world, firm-designed, non-time-based budget (NTB) model pilot program during which several engagement teams were not provided with a time budget and auditors did not track time. Results suggest auditors appreciated the autonomy the NTB model allows, felt it improved teamwork, reported less concern about performance evaluations, felt empowered to deliver value-added services to the client, and indicated greater work-life balance and job satisfaction. Firm-provided proprietary archival evidence suggests that efficiency did not differ across the two models, as the timing of audit procedure signoffs was not negatively impacted under the NTB model. Overall, our results suggest a NTB model may be a viable option for audit firms looking to decrease their reliance on time-based budgeting. Data Availability: Please contact Erin M. Hawkins.

Time-Based Rational Function Model for Block Adjustment of Long-Strip LROC NAC Images Using Sparse Elevation Controls

Time-Based Rational Function Model for Block Adjustment of Long-Strip LROC NAC Images Using Sparse Elevation Controls

Certain Investigations on Feature Selection Technique Using Artificial Immune Systems for EEG Color Visualization Classification

This paper aims to extract and select the significant features of electroencephalogram (EEG) signals and classify the visual stimulation of distinct colors. In this work, a novel method for selecting distinct colors using EEG signals called affinity artificial immune and Daubechies wavelet time-based learning (AAIDWTL) is proposed. Initially, the EEG signals were collected in a controlled environment and an in-built band-pass filter was applied to remove the artifacts. The filtered signals were converted into frequency domain signals using least squarebased short-term Fourier transform. After that, by utilizing Daubechies wavelet statistical time-based feature extraction model the time domain features were extracted. Followed by, computationally efficient features were selected using an affinity artificial immune-based feature selection model. The selected features were classified using a polynomial kernel multiclass classification-based machine learning algorithm and achieved an accuracy of 97.5% when compared with other methods like linear discriminant analysis (LDA) which obtained only 92%. Furthermore, while utilizing the proposed method classification time was considerably less when compared to LDA. The experimental result shows that the proposed color stimulation of the EEG signals method achieved greater improvement in terms of both classification time and classification accuracy with a minimum false positive rate.

Calculation of Power Purchase Imbalance Costs for Grid Enterprises’ Agents Considering Uncertainty of Supply and Demand and Price Fluctuations in the Context of Beijing-Tianjin-Tangshan Unified Spot

Abstract The regional unified electricity spot market plays an important role in promoting the optimal allocation of power resources over a wider range and cross-regional consumption of renewable energy. This paper examines the supply and demand uncertainty and price volatility of power grid agency power purchases in the context of the Beijing-Tianjin-Tangshan unified spot purchase. It delves into the specific mechanism of unbalanced costs in agency power purchase decisions and investigates the specific channeling methods of unbalanced costs in the dual-track system. Based on the dual-track unbalanced cost under the unified spot market, a dual-track power balance model is proposed based on the typical daily load curve, and a time-based unbalanced cost calculation model is designed. Using the typical daily electricity load data from the power supply enterprise in X province as an example, the calculation and simulation of unbalanced costs are carried out. The spot electricity price fluctuates between 0.25 yuan/(kW•h) and 0.45 yuan/(kW•h) in different periods. The overall unbalanced power purchased by the power grid in the load trough stage is only 6.24% of the total monthly electricity consumption. Still, the unbalanced cost in the study period is of a deficit nature, and its average value reaches −268 million yuan. The calculation method of unbalanced costs caused by agency power purchases clearly reveals the unbalanced cost-sharing mechanism in the context of the unified spot.

From principles to practice: Implementation of entrustable professional activities (EPAs) for surgical pathology residency education in a large academic hospital

Over the past decade, competency-based medical education (CBME) has gained momentum in the United States to develop trainees into independent and confident physicians by the end of their training. Entrustable professional activities (EPAs) are an established methodology for assessing trainee development through an outcomes-driven rather than a time-based model. While EPAs have been utilized as an assessment tool for CBME in Europe and Canada, their validation and implementation in some medical specialties has occurred more recently in the United States. Pediatrics was the first specialty in the US to conduct a large-scale UME-GME pilot. Pathology Residency EPAs were published in 2018; however, implementation in training programs has been slow. We have piloted EPAs in our residency program's surgical pathology rotation and propose a unique set of 4 surgical pathology EPAs to track trainee preparedness for independent practice.

Semiparametric multivariate joint model for skewed-longitudinal and survival data: A Bayesian approach.

Joint models and statistical inference for longitudinal and survival data have been an active area of statistical research and have mostly coupled a longitudinal biomarker-based mixed-effects model with normal distribution and an event time-based survival model. In practice, however, the following issues may standout: (i) Normality of model error in longitudinal models is a routine assumption, but it may be unrealistically violating data features of subject variations. (ii) Data collected are often featured by the mixed types of multiple longitudinal outcomes which are significantly correlated, ignoring their correlation may lead to biased estimation. Additionally, a parametric model specification may be inflexible to capture the complicated patterns of longitudinal data. (iii) Missing observations in the longitudinal data are often encountered; the missing measures are likely to be informative (nonignorable) and ignoring this phenomenon may result in inaccurate inference. Multilevel item response theory (MLIRT) models have been increasingly used to analyze the multiple longitudinal data of mixed types (ie, continuous and categorical) in clinical studies. In this article, we develop an MLIRT-based semiparametric joint model with skew-t distribution that consists of an extended MLIRT model for the mixed types of multiple longitudinal data and a Cox proportional hazards model, linked through random-effects. A Bayesian approach is employed for joint modeling. Simulation studies are conducted to assess performance of the proposed models and method. A real example from primary biliary cirrhosis clinical study is analyzed to estimate parameters in the joint model and also evaluate sensitivity of parameter estimates for various plausible nonignorable missing data mechanisms.

Phenological development of subterranean clover (Trifolium subterraneum L.) in New Zealand in response to different climate change scenarios

Climate predictions for New Zealand for the coming decades suggest rising temperatures (+0.7–3 °C), increased diurnal temperature and variable precipitation patterns that differ around the country and with seasons. The most common pattern of annual precipitation indicates the largest increases in the west of the South Island and the largest decreases in the east of the North Island and coastal Marlborough. The phenophases of subterranean clover were estimated and quantified using a thermal time-based model under different climate scenarios based on greenhouse gases and aerosol pathways over the 21st century, known as the Representative Concentration Pathways (RCPs), at two periods (mid and end of the century). The estimated flowering (R3) and post- flowering (R6-R11) date change showed a consistent trend across all districts. The largest date advances (≥ 5 days) were predicted for the three current latest flowering districts (Mackenzie, Queenstown and Central Otago). A reduction of the plant’s life cycle, for both ‘Early’ and ‘Late’ maturity cultivars may have undesired consequences for forage yield so genetic material with later flowering dates may be useful. Adaptation strategies to mitigate the future warming effects include using ‘Late’ flowering cultivars and greater use of supplementary feed through summer dry conditions.

Comparison of the response time-based effort-moderated IRT model and three-parameter logistic model according to computerized adaptive test performances: a simulation study

Depending on the developments in technology and information, paper-pencil tests leave their place for computerized adaptive tests (CATs). CAT is widely used in the field of health, mainly in psychiatry. Many item response theory models have been proposed in the literature regarding the use of response time focusing on item difficulty and personal characteristics by ignoring the multidimensional interactions, therefore these results may cause bias in estimates of individual ability levels. The present simulation study was conducted to compare the performance of CAT applications of the effort-moderated item response theory (EM-IRT) model, which is based on response time, and the three-parameter logistic (3PL) model. While simulating CAT with the EM-IRT model and the 3PL model, the hybrid method was used for ability estimation and maximum Fisher information (MFI) was used for item selection. The CAT process proceeded until the standard error of the estimation was <0.3 and <0.5, or all items in the item bank were used. The number of individuals was specified as 1000, while the number of items was changed to 50, 100, and 250. All six scenarios were repeated 1000 times. With the increase in the number of items and the decrease in the standard error as a stopping criterion, consistent results were obtained with true ability levels in both methods. The CAT with the EM-IRT model estimated true ability level slightly lower than CAT with the 3PL model. The EM-IRT model enables measuring the response time that could yield additional data to the physician about the mental and cognitive condition of the patient. The CAT method can be a promising method of telemedicine in the era of the pandemic.

Analysis of PAUD Learning Model Beyond Centers and Circle Time (BCCT) and Its Implementation In Educational Institutions

Early Childhood Education (PAUD) is an important stage for a child's growth. The selection of educational institutions with certain learning models, it is important for parents to consider before making a choice in one school. There are various kinds of PAUD learning models. Among these learning models are the classical model, the cooperative learning model, the center and circle time-based learning model (beyond centers and circle time/BCCT) and the area learning model (interest). Among these models, the center-based learning model and circle time are learning models that are still rarely applied in PAUD. This study aims to analyze the center-based learning model and circle time in terms of preparation and implementation. The approach used in this study is a qualitative approach. Data were obtained through interview, observation and various literature related to PAUD learning models. The results show that the center-based learning model and circle time is a learning model with four steps, namely the environmental footing, the footing before playing, the footing during playing and the footing after playing in the learning activities. The application of this learning model requires teachers to carry out preparatory activities every day. It is hoped that the aspects of children's development can improve through various series of activities in the central class which are followed by children alternately every day.

Determination of Distribution Center Location using Analysis of Time-Based Set Covering Model and Maximal Covering Model Analysis

The distribution or delivery process is one factor that affect customer satisfaction as the goal of supply chain. In order for supply chain to be competitive in competition, delivery time is an important factor to manage, so that it could provide high service level value. One factor that could affects distribution or delivery is distribution center location. A strategic distribution center location would facilitate and speed up distribution process. This research discusses model for determining location of distribution center so that delivery could be made at the right amount and in the right time by analyzing of time-based using set covering model and maximal covering model. The numerical example in this research is case study on paint supply chain in Bandung which has a number of delivering clusters. Each cluster consists of a number of building material stores which are located scattered. There is one cluster that has problems in fulfilling demands and delivery times. The analysis made could be used to determine the distribution center location so that it could serve all consumer demands. Based on the calculation results, it was found that there were two distribution center locations. Two distribution centers could fulfill the objective function of cost minimization and satisfy all demands. The results of the sensitivity analysis show that there are several factors that influence the determination of location number.

A probe into the fatigue crack growth in mechanical systems with hyperchaotic/chaotic dynamics

This study presents experimental and analytical fatigue frameworks for the dynamic fatigue life prediction of nonlinear mechanical systems under hyperchaotic oscillations. Fatigue in aluminium notched beams under hyperchaotic/chaotic vibrations is studied using a time-based analytical model. The fatigue lives obtained by the analytical model provided satisfactory correlations with those acquired by the dynamic fatigue experiments conducted on cracked aluminium beams. The model is then used to study the fatigue under Duffing chaotic, Duffing hyperchaotic, and periodic dynamic forcing. Results indicate that the classic linear cumulative damage rule delivers inaccurate high cycle fatigue (HCF) predictions under chaotic and hyperchaotic loading. Proper prediction of the HCF in metallic mechanical components experiencing chaotic and hyperchaotic dynamics requires time-based fracture-mechanics analytical models, such as that proposed in the current study. Under similar elastic energy inputs, hyperchaotic loading brings higher fatigue damage than chaotic and periodic loading. The fatigue life sensitivity to initial conditions is high under hyperchaotic, moderate under chaotic and next-zero under periodic loading.

Robust Federated Learning for execution time-based device model identification under label-flipping attack

The computing device deployment explosion experienced in recent years, motivated by the advances of technologies such as Internet-of-Things (IoT) and 5G, has led to a global scenario with increasing cybersecurity risks and threats. Among them, device spoofing and impersonation cyberattacks stand out due to their impact and, usually, low complexity required to be launched. To solve this issue, several solutions have emerged to identify device models and types based on the combination of behavioral fingerprinting and Machine/Deep Learning (ML/DL) techniques. However, these solutions are not appropriate for scenarios where data privacy and protection are a must, as they require data centralization for processing. In this context, newer approaches such as Federated Learning (FL) have not been fully explored yet, especially when malicious clients are present in the scenario setup. The present work analyzes and compares the device model identification performance of a centralized DL model with an FL one while using execution time-based events. For experimental purposes, a dataset containing execution-time features of 55 Raspberry Pis belonging to four different models has been collected and published. Using this dataset, the proposed solution achieved 0.9999 accuracy in both setups, centralized and federated, showing no performance decrease while preserving data privacy. Later, the impact of a label-flipping attack during the federated model training is evaluated using several aggregation mechanisms as countermeasures. Zeno and coordinate-wise median aggregation show the best performance, although their performance greatly degrades when the percentage of fully malicious clients (all training samples poisoned) grows over 50%.

WI-FI-BASED CHANNEL STATE INFORMATION APPROACH FOR SENSING AND HUMAN WALKING ACTIVITY RECOGNITION IN INDOOR ENVIRONMENT

Many studies have demonstrated that Wireless Fidelity (Wi-Fi) networks can detect human behavior due to the development of wireless devices and detecting methodologies. Human activity recognition applied Channel State Information (CSI) in commercial Wi-Fi devices, which has many practical applications. This paper aims to present a method for recognizing human activity that uses enhanced Wi-Fi-based CSI signals. At first, researchers examined how various perceptive antennas detect motion caused by humans. The proposed algorithm for selecting antennas considers the sensitivity of each antenna concerning the task at hand. Long short-term memory (LSTM) and Inception Time-based classification model was developed and fine-tuned for experiments. This study also suggested two basic power methods to retrieve the effective interval associated with human activity, one of which can solidify the active signals while weakening the inactive signals. Last but not least, an algorithm has been developed for segmenting activity to identify initiation and termination times. An efficient outliers removal technique was used to remove the edges. The results show that using commonplace Wi-Fi routers can accurately categorize human actions 87% of the time. To validate and evaluate these methods, immense experiments were conducted in exquisite indoor environments. As demonstrated by experiments, the proposed methods can mitigate the disturbance caused by inert and non-responsive signals and enhance accuracy rates while improving efficiency and processing time.

A Comprehensive Overview of Additive Manufacturing Processes Through a Time-Based Classification Model.

The ongoing crisis caused by the COVID-19 pandemic produced major reshuffles on the world map, bringing imbalance, uncertainty, and accumulated stress. Due to supply chain disruptions, the need for innovation has emerged both as a priority and a necessity and three-dimensional printing (3DP) proved to be a primary, smart, effective, and innovative additive manufacturing (AM) method. AM refers to the direct fabrication of complex geometries, using a computer-aided design (CAD) model or a three-dimensional scanner output. This article presents a literature review of AM technologies, chronologically sorted, and proposes a multilevel classification model. The suggested research approach appears a triangular methodology that encompasses the current ISO/ASTM 52900:2021 report. The first objective of this article is to form two double-level classification models of AM processes, depending on the technology and material factors. The second objective is to clarify in which of the proposed categories each AM process is included; and the third one is to investigate if the proposed taxonomy is related to the time spot, in which AM processes were invented. The contribution of this article lies in determining the factors that are crucial for the growth of AM ecosystem. The novelty of the proposed classification lies in the definition of an optimal option for each industrial application based on the different AM processes, the variety of materials, and the evolution of technology over the years. In this way, investing in AM is more systematic and less risky.

Neural Correlates of Cognitive Load While Playing an Emergency Simulation Game: A Functional Near-Infrared Spectroscopy (fNIRS) Study

Functional near-infrared spectroscopy (fNIRS) provides reliable results for determining cognitive load based on averaged cortical blood flow during multiple repetitions of short cognitive tasks. At the same time, it remains unclear how to use this technique for assessing cognitive load during prolonged single-trial activity. In this study, we used a computer-based emergency simulation game for inducing different levels of cognitive load. We propose a novel approach to measure cognitive load using specific time slots, determined based on simulation log-data interpreted in light of Barrouillet's time-based resource-sharing model. To validate this approach, we compared cortical activity in dorsolateral prefrontal cortex (DLPFC) and left inferior frontal gyrus (IFG) regions measured at four specific time slots during a simulation. We found significant associations between cognitive load and neuronal activity within the DLPFC depending on the chosen time slot, whereas no such dependencies were found for the IFG. These results illustrate how knowledge of task structure could be used advantageously for the identification of cognitive load. Although requiring further investigation in terms of reliability and generalizability, the presented approach can be considered promising evidence that fNIRS might be suitable for more general reliable assessments of cognitive load during prolonged single-trial activities and for real-time adaptations in simulation-based learning environments.

Theory-based approach for assessing cognitive load during time-critical resource-managing human–computer interactions: an eye-tracking study

Computerized systems are taking on increasingly complex tasks. Consequently, monitoring automated computerized systems is becoming increasingly demanding for human operators, which is particularly relevant in time-critical situations. A possible solution might be adapting human–computer interfaces (HCI) to the operators’ cognitive load. Here, we present a novel approach for theory-based measurement of cognitive load based on tracking eye movements of 42 participants while playing a serious game simulating time-critical situations that required resource management at different levels of difficulty. Gaze data was collected within narrow time periods, calculated based on log data interpreted in the light of the time-based resource-sharing model. Our results indicated that eye fixation frequency, saccadic rate, and pupil diameter significantly predicted task difficulty, while performance was best predicted by eye fixation frequency. Subjectively perceived cognitive load was significantly associated with the rate of microsaccades. Moreover our results indicated that more successful players tended to use breaks in gameplay to actively monitor the scene, while players who use these times to rest are more likely to fail the level. The presented approach seems promising for measuring cognitive load in realistic situations, considering adaptation of HCI.