Exploring a Bilingual Next Word Predictor for a Federated Learning Mobile Application

We explore a bilingual next-word predictor (NWP) under federated optimization for a mobile application. A character-based LSTM is server-trained on English and Dutch texts from a custom parallel corpora. This is used as the target performance. We simulate a federated learning environment to assess the feasibility of distributed training for the same model. The popular Federated Averaging (FedAvg) algorithm is used as the aggregation method. We show that the federated LSTM achieves decent performance, yet it is still sub-optimal. We suggest possible next steps to bridge this performance gap. Furthermore, we explore the effects of language imbalance varying the ratio of English and Dutch training texts (or clients). We show the model upholds performance (of the balanced case) up and until a 80/20 imbalance before decaying rapidly. Lastly, we describe the implementation of local client training, word prediction and client-server communication in a custom virtual keyboard for Android platforms. Additionally, homomorphic encryption is applied to provide with secure aggregation guarding the user from malicious servers.

Exploring a Bilingual Next Word Predictor for a Federated Learning Mobile Application

We explore a bilingual next-word predictor (NWP) under federated optimization for a mobile application. A character-based LSTM is server-trained on English and Dutch texts from a custom parallel corpora. This is used as the target performance. We simulate a federated learning environment to assess the feasibility of distributed training for the same model. The popular Federated Averaging (FedAvg) algorithm is used as the aggregation method. We show that the federated LSTM achieves decent performance, yet it is still sub-optimal. We suggest possible next steps to bridge this performance gap. Furthermore, we explore the effects of language imbalance varying the ratio of English and Dutch training texts (or clients). We show the model upholds performance (of the balanced case) up and until a 80/20 imbalance before decaying rapidly. Lastly, we describe the implementation of local client training, word prediction and client-server communication in a custom virtual keyboard for Android platforms. Additionally, homomorphic encryption is applied to provide with secure aggregation guarding the user from malicious servers.

RT-PCR/MALDI-TOF diagnostic target performance reflects circulating SARS-CoV-2 variant diversity in New York City

As severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to circulate, multiple variants of concern (VOC) have emerged. New variants pose challenges for diagnostic platforms since sequence diversity can alter primer/probe binding sites (PBS), causing false-negative results. The Agena MassARRAY® SARS-CoV-2 Panel utilizes reverse-transcription polymerase chain reaction and mass-spectrometry to detect five multiplex targets across N and ORF1ab genes. Herein, we utilize a dataset of 256 SARS-CoV-2-positive specimens collected between April 11, 2021-August 28, 2021 to evaluate target performance with paired sequencing data. During this timeframe, two targets in the N gene (N2, N3) were subject to the greatest sequence diversity. In specimens with N3 dropout, 69% harbored the Alpha-specific A28095U polymorphism that introduces a 3’-mismatch to the N3 forward PBS and increases risk of target dropout relative to specimens with 28095A (relative risk (RR): 20.02; p<0.0001; 95% Confidence Interval (CI): 11.36-35.72). Furthermore, among specimens with N2 dropout, 90% harbored the Delta-specific G28916U polymorphism that creates a 3’-mismatch to the N2 probe PBS and increases target dropout risk (RR: 11.92; p<0.0001; 95% CI: 8.17-14.06). These findings highlight the robust capability of Agena MassARRAY® SARS-CoV-2 Panel target results to reveal circulating virus diversity and underscore the power of multi-target design to capture VOC.

Performance Analysis of IRS-Aided NOMA Communications in the Presence of Imperfect SIC

The intelligent reflecting surface (IRS) is expected to be a promising technique to achieve a robust spectrum and energy efficiency. This paper investigates the advantages of IRS in enhancing performance of non-orthogonal multiple access (NOMA) communications in the presence of imperfect successive-interference-cancellation (SIC) and phase distortion (PD) caused by a non-ideal IRS. Specifically, average achievable rates (AARs) of the users are the target performance metrics. For performance evaluation, the probabilistic characterizations of signal-to-interference-plus-noise ratios (SINRs) at the users are studied. These results allow for deriving the theoretical formulas for the AAR. Monte Carlo simulations are adopted to verify the accuracy of these theoretical results. The numerical results show the effects of various key system parameters, such as source transmit power, NOMA power allocation (PA) factors, reflecting tile (RTs) allocation, the SIC imperfection factor, and the PD factor, on the AAR that provide useful information for the system’s design.

Robust Explorative Particle Swarm Optimization for Optimal Design of EV Traction Motor

This paper proposes a robust optimization algorithm customized for the optimal design of electric machines. The proposed algorithm, termed “robust explorative particle swarm optimization” (RePSO), is a hybrid algorithm that affords high accuracy and a high search speed when determining robust optimal solutions. To ensure the robustness of the determined optimal solution, RePSO employs the rate of change of the cost function. When this rate is high, the cost function appears as a steep curve, indicating low robustness; in contrast, when the rate is low, the cost function takes the form of a gradual curve, indicating high robustness. For verification, the performance of the proposed algorithm was compared with those of the conventional methods of robust particle swarm optimization and explorative particle swarm optimization with a Gaussian basis test function. The target performance of the traction motor for the optimal design was derived using a simulation of vehicle driving performance. Based on the simulation results, the target performance of the traction motor requires a maximum torque and power of 294 Nm and 88 kW, respectively. The base model, an 8-pole 72-slot permanent magnet synchronous machine, was designed considering the target performance. Accordingly, an optimal design was realized using the proposed algorithm. The cost function for this optimal design was selected such that the torque ripple, total harmonic distortion of back-electromotive force, and cogging torque were minimized. Finally, experiments were performed on the manufactured optimal model. The robustness and effectiveness of the proposed algorithm were validated by comparing the analytical and experimental results.

The role of capabilities in shaping sustainable supply chain flexibility and enhancing circular economy-target performance: an empirical study

Purpose A circular economy is a popular approach considered by many firms to address sustainable development goals strategically. Literature indicates that collaborative relationships among supply chain partners facilitate circular economy practices. However, there is a dearth of studies in lower-middle-income countries indicating the unique challenges industries face whilst practising circular economy principles and how the challenges can be overcome. To address the calls of previous researchers, this study aims to explore the following relationships: engagement and alliance capability whilst data analytics capability plays a mediating role; the relationship between alliance and data analytics capability with sustainable supply chain flexibility whilst industry dynamism is considered as a moderating variable and the relationship between sustainable supply chain flexibility and circular economy-target performance. Design/methodology/approach A cross-sectional survey was performed and data was collected from 760 employees of Indian firms. Covariance-based structural equation modelling was applied to perform the path analysis to determine a firm’s capabilities in shaping sustainable supply chain flexibility and enhancing circular economy target performance. Findings Drawing upon dynamic capability theory, it was first established that engagement capability has a positive and significant influence on alliance capability, whilst data analytics capability played a partial mediating role. Second, it was established that alliance capability and data analytics capability significantly affect sustainable supply chain flexibility, whilst industry dynamism played a moderating role. Finally, it was clear that sustainable supply chain flexibility had a significant and positive effect on circular economy target performance, ultimately enhancing sustainability. Originality/value This study advances the circular economy literature by recommending that firms must consider some critical operational level capabilities to develop their dynamic capability, i.e. sustainable supply chain flexibility, to better meet the competitive market conditions in turbulent business environments.

Deep Learning Model Explainability for Inspection Accuracy Improvement in the Automotive Industry

The welding seams visual inspection is still manually operated by humans in different companies, so the result of the test is still highly subjective and expensive. At present, the integration of deep learning methods for welds classification is a research focus in engineering applications. This work intends to apprehend and emphasize the contribution of deep learning model explainability to the improvement of welding seams classification accuracy and reliability, two of the various metrics affecting the production lines and cost in the automotive industry. For this purpose, we implement a novel hybrid method that relies on combining the model prediction scores and visual explanation heatmap of the model in order to make a more accurate classification of welding seam defects and improve both its performance and its reliability. The results show that the hybrid model performance is relatively above our target performance and helps to increase the accuracy by at least 18%, which presents new perspectives to the developments of deep Learning explainability and interpretability.

THE STYLE OF THINKING ACTIVITY OF MODERN YOUTH AS A SCIENTIFIC AND PRACTICAL PROBLEM

The statistic is based on the state of the art of scientific advancement in the style of misinformation of the modern student youth. The style of the target is to be seen as a system of intellectual strategies, priyom, a quick reference to the operation of a specialty, how to be victorious with it in a target performance. To respect the links to the style of the targeting from the cognitive style as the way of receiving and processing information. It is to be determined that the effectiveness of the educational performance of students is significant because of the fact that the methodology of victoriousness will be the same because of the dominant style characteristics of their thinking acyivity. Respect for class is characteristic of the specialty of the style of thinking of the current student youth. It is emphasized that the tendency to reproduce heterogeneous objects, without coherent elements between them, which are characterized by illogicality, heterogeneity and high speed of switching between pieces of information, leads to difficulties in forming the desired quality of students' thinking – criticality as the ability to assess reliability, accuracy or values of something, the ability to search for and find causes and alternative points of view, to perceive the situation as a whole or to change one's position on the basis of facts and arguments. It is noted that in order to optimize the educational process in university it is necessary to select such teaching methods that would ensure the effectiveness of the thinking process of students.

Gaze signals both cue direction and agent mental states

Although it is well established that humans spontaneously follow where others are looking, it remains debated if this gaze following behaviour occurs due to the gaze cue’sdirectional information (i.e., where an agent is attending) or the agent’s inferred mental state (i.e., what they are attending to). We tested this notion by assessing the combined anddissociated effects of the gaze cue direction and the agent’s mental content. Gazed-at target performance was compromised when cue direction and inferred mental content were dissociated relative to when they were combined. This effect was especially prominent for social relative to nonsocial cues. Thus, gaze signals include information about both the cue direction and the gazer’s mental content, communicating information about where gazed at items are located and what those items are.