Enhancing Transformer Training Efficiency with Dynamic Dropout

Iterative multiuser detection has gained increasing attention, as a powerful method to mitigate the effects of multiple access interference (MAI) in CDMA systems. The polynomial expansion (PE) multiuser detector discussed in this paper can also be seen as an iterative approach to approximate the MMSE multiuser detector. It is shown that the PE detector can be applied such that it has fast convergence (few iterations) and high performance (comparable to the MMSE multiuser detector), while its computational complexity O(N/sup 2/) remains well below that of most multiuser detectors. Simulation results for the third generation mobile communication system UTRA-TDD show the viability of the detector. Even in a time-variant multipath channel the PE detector can sustain its properties. The fast convergence of the PE detector is due to a very close estimation of the eigenvalues of the correlation matrix.

A facile method was developed to produce hemiacetal ester DCPNs combining fast reprocessability and high performance.

This study compared popular Deep Learning (DL) architectures to classify machining surface roughness using sound and force data. The DL architectures considered in this study include Multi-Layer Perceptron (MLP), Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), and transformer. The classification was performed on the sound and force data generated during machining aluminum sheets for different levels of spindle speed, feed rate, depth of cut, and end-mill diameter, and it was trained on 30 s machining data (10–40 s) of the machining experiments. Since a raw audio waveform is seldom used in DL models, Mel-Spectrogram and Mel Frequency Cepstral Coefficients (MFCCs) audio feature extraction techniques were used in the DL models. The results of DL models were compared for the training–validation accuracy, training epochs, and training parameters of each model. Although the roughness classification by all the DL models was satisfactory (except for CNN with Mel-Spectrogram), the transformer-based modes had the highest training (>96%) and validation accuracies (≈90%). The CNN model with Mel-Spectrogram exhibited the worst training and inference accuracy, which is influenced by limited training data. Confusion matrices were plotted to observe the classification accuracy visually. The confusion matrices showed that the transformer model trained on Mel-Spectrogram and the transformer model trained on MFCCs correctly predicted 366 (or 91.5%) and 371 (or 92.7%) out of 400 test samples. This study also highlights the suitability and superiority of the transformer model for time series sound and force data and over other DL models.

PurposeScale-ups have a crucial role in our society and economy, are known for their fast growth and high performance and undergo significant organizational change. Research on the organizational elements that ensure scale-ups sustaining high performance is limited. This empirical study aims to investigate the organizational culture in scale-ups using the Competing Values Framework, including the clan, adhocracy, the market, hierarchy cultures and its relation to performance.Design/methodology/approachQuantitative and qualitative data were collected in five scale-ups. Surveys provided data of 116 employees on organizational culture, assessed using the Organizational Culture Assessment Instrument (OCAI) and perceived performance. The aggregate outcomes and performance measures were analyzed using correlation analysis. Interviews were held with ten top managers and mirrored against the quantitative data.FindingsThe results show that top managers and employees have different perspectives on the culture scale-ups are supposed to have. Top managers perceive market culture as more and hierarchy culture as less present in their organizations than employees. The clan and adhocracy culture are positively correlated to performance and are preferred by employees. Market and hierarchy culture types are negatively correlated to performance and are least preferred by employees.Research limitations/implicationsIt enables scale-up leaders to specifically intervene when cultural elements are experienced by employees that will not drive performance and fit the high performance and fast-growing scale-up environment.Originality/valueThis study is the first that showed that organizational-wide surveys, combined with in-depth interviews, are suitable for top managers of scale-ups to diagnose the organizational culture and the effect on the organization's performance.

The need for low-cost high-performance broadband photon detection with sensitivity in the near infrared (NIR) has driven interest in new materials that combine high absorption with traditional electronic infrastructure (CMOS) compatibility. Here, we demonstrate a facile, low-cost and scalable, catalyst-free one-step solution-processed approach to grow one-dimensional Sb2Se3 nanostructures directly on flexible substrates for high-performance NIR photodetectors. Structural characterization and compositional analyses reveal high-quality single-crystalline material with orthorhombic crystal structure and a near-stoichiometric Sb/Se atomic ratio. We measure a direct band gap of 1.12 eV, which is consistent with predictions from theoretical simulations, indicating strong NIR potential. The fabricated metal-semiconductor-metal photodetectors exhibit fast response (on the order of milliseconds) and high performance (responsivity ~ 0.27 A/W) as well as excellent mechanical flexibility and durability. The results demonstrate the potential of molecular-ink-based Sb2Se3 nanostructures for flexible electronic and broadband optoelectronic device applications.

Introduction. This study introduces a sliding mode control (SMC) that utilizes multivariable system command estimation (MSCE-SMC) to create an innovative speed control system for the permanent magnet synchronous motor (PMSM). The motor operates through a 3-phase voltage source inverter when used in an electric vehicle (EV) model, with the goal of achieving fast speed regulation and high performance. Problem. The primary challenge is to achieve fast and accurate speed regulation for PMSMs while maintaining high performance, despite varying system parameters and external disturbances. The goal is to design a robust and adaptive speed control system for PMSMs using the SMC approach, which ensures precise speed tracking and high-performance regulation. Scientific novelty. The integration of MSCE-SMC approach, offering an innovative solution for speed control in PMSMs used in EVs. Methodology. SMC approach for the PMSM divides the system into 2 subsystems: electrical and speed. A d-q coordinate frame is used to model the PMSM, and its control strategy is outlined. A detailed model of the PMSM with SMC is presented after an in-depth review of the theoretical concepts and principles of sliding mode control. Results. To validate the proposed approach, MATLAB/Simulink is conducted, demonstrating the effectiveness and robustness of the method in PMSM speed regulation. The results confirm that the proposed method provides straightforward and precise control, accurate speed tracking, and high-performance regulation. It also shows adaptability to parameter variations and external disturbances. Practical value. The practical value of the proposed method is significant, as it provides a reliable and efficient control system for PMSMs. It offers precise speed control, robust performance under variable conditions, and high adaptability to external disturbances, making it suitable for real-world EV applications. References 22, table 1, figures 18.

This paper presents an analytical investigation into the contraction of spherical and cylindrical cavities excavated in strain-softening rock masses obeying the Mohr–Coulomb or Tresca yield criterion, with linear or exponential uniaxial compressive strength decay. The derivation of the ground response curves is based on the simplifying assumption that the strains inside the plastic zone are completely plastic. This significantly simplifies the mathematical formulation, enabling the derivation of closed-form solutions. An alternative simplifying approach which partially neglects the elastic strains inside the plastic zone and which is commonly adopted in the literature, is also examined. The accuracy of the simplified solutions is evaluated by comparing their predictions with rigorous solutions obtained by numerical finite-difference analyses. The investigation demonstrates that the proposed closed-form solutions represent a significant improvement on those based on the commonly-made simplifying assumption involving partial neglect of elastic strains.

In this paper, a modified Bounding Ellipsoidal Adaptive CONstrained least-squares (M-BEACON) algorithm is introduced that provides fast convergence performance by using short training sequences in stationary and quasi stationary environments. The fast convergence performance of the modified algorithm depends on the selection of forgetting factor and scaling parameter depends upon the Doppler rate and the set membership constraint, respectively. Simulation results show that M-BEACON algorithm provides fast convergence than the conventional BEACON and RLS algorithms. Moreover, the update rate of the modified algorithm is much lesser than that of BEACON algorithm which makes the critical impact on the computational cost.

Abstract. Crowd counting, a critical component in the management and safety planning of large gatherings and public spaces, is essential for ensuring smooth event operations and preventing potential overcrowding issues. While the standard convolutional neural network (CNN) based model performs well in head counting tasks, it has certain drawbacks when applied to complex scenarios. With the rapid development of artificial intelligence, Transformer models that rely on self-attention mechanisms, as Swin Transformer, have demonstrated exceptional performance in visual tasks, such as image classification, and segmentation in recent times. This study examines the experimental findings of Swin Transformer's head counting tasks and contrasts them with the CNN-based model. Mean Absolute Error (MAE) and Mean Square Error (MSE) evaluation indicators show that the Transformer model outperforms the classic CNN model in terms of generalization abilities when dealing with complicated scenarios. Future research work will increase the diversity of data sets and focus on optimizing model structure and improving training efficiency.

Dynamic alterations in flavin adenine dinucleotide (FAD) fluorescence permit insight into energy metabolism-dependent changes of intramitochondrial redox potential. Monitoring FAD fluorescence in living tissue is impeded by photobleaching, restricting the length of microfluorimetric recordings. In addition, photodecomposition of these essential electron carriers negatively interferes with energy metabolism and viability of the biological specimen. Taking advantage of pulsed LED illumination, here we determined the optimal excitation settings giving the largest fluorescence yield with the lowest photobleaching and interference with metabolism in hippocampal brain slices. The effects of FAD bleaching on energy metabolism and viability were studied by monitoring tissue pO2 , field potentials and changes in extracellular potassium concentration ([K+ ]o ). Photobleaching with continuous illumination consisted of an initial exponential decrease followed by a nearly linear decay. The exponential decay was significantly decelerated with pulsed illumination. Pulse length of 5 ms was sufficient to reach a fluorescence output comparable to continuous illumination, whereas further increasing duration increased photobleaching. Similarly, photobleaching increased with shortening of the interpulse interval. Photobleaching was partially reversible indicating the existence of a transient nonfluorescent flavin derivative. Pulsed illumination decreased FAD photodecomposition, improved slice viability and reproducibility of stimulus-induced FAD, field potential, [K+ ]o and pO2 changes as compared to continuous illumination.

Diabetic retinopathy detection through convolutional neural networks with synaptic metaplasticity

Poor compatibility and sluggish kinetics at the hard carbon electrode-electrolyte interface represent a critical bottleneck limiting the development of high-performance sodium-ion batteries. To address this issue, this study introduces Zn(OTf)2 and NaSO2CF3 as multifunctional composite additives, constructing an electrolyte system capable of both interface modification and capacity compensation. Experimental results demonstrate that Zn(OTf)2 preferentially decomposes and in situ forms an SEI film rich in NaZn13 on the hard carbon surface. This NaZn13 component exhibits excellent stability and sodium-ion conductivity, promoting uniform Na+ distribution and rapid migration, thereby enhancing interfacial kinetics. Concurrently, NaSO2CF3 acts as a sodium reservoir that can fully decompose during cycling, effectively compensating for the irreversible capacity loss in the first cycle. Based on this synergistic strategy, the initial Coulombic efficiency of the Na4Fe3(PO4)P2O7||HC full cell increases from 67.62% to 80.78%, and the capacity retention after 1000 cycles improves from 56.08% to 86.93%. This design not only modifies the interface but also compensates for the active sodium loss incurred during the alloying process, significantly enhancing the interfacial performance and cycling stability of the battery and offering promising application prospects.

Fifth-generation (5G) communications system is commercially introduced by several mobile operators where sub-6 GHz bands are the backbone of the 5G networks. A large-scale multiple-input multiple-output (MIMO), or massive MIMO (mMIMO), technology has a major impact to secure high data rate, high spectral efficiency, and quality of service (QoS). It could also have a major role in the beyond-5G systems. A massive number of antennas seek advanced signal processing to detect and equalize the signal. However, optimal detectors, such as the maximum likelihood (ML) and maximum posterior (MAP), are not desirable in implementation due to extremely high complexity. Therefore, sub-optimum solutions have been introduced to obtain and guarantee enough balance between the performance and the computational complexity. In this paper, a robust and joint low complexity detection algorithm is proposed based on the Jacobi (JA) and Gauss–Seidel (GS) methods. In such iterative methods, the performance, complexity, and convergence rate are highly dependent on the initial vector. In this paper, initial solution is proposed by exploiting the benefits of a stair matrix to obtain a fast convergence rate, high performance, and low complexity. Numerical results show that proposed algorithm achieves high accuracy and relieve the computational complexity even when the BS-to-user-antenna ratio (BUAR) is small.

We investigated the effect of dielectric relaxation in polymer gate insulators on the device characteristics of organic field effect transistors. Dielectric relaxation of polymer gate insulators caused an increase in drain current (ID) in a period starting immediately after the application of the gate voltage (VG) and lasting several milliseconds. This induced an apparent delay in the response of ID. Based on the observed results, we suggested an ideal gate insulator to achieve organic field effect transistors that have a fast response and high performance.

Envelope tracking has many important applications in electrical distribution systems for either flicker meters or flicker compensators. This paper presents a novel control technique for flicker mitigation. This technique is based on the instantaneous tracking of the measured voltage envelope. The ADAptive LINEar neuron (ADALINE) algorithm, and the Recursive Least Square (RLS) algorithm are introduced for the flicker envelope tracking. The ADALINE algorithm is characterized by a light computational demand, unlike that of existing techniques; the RLS algorithm outperforms the ADALINE algorithm by its fast convergence and robust tracking performance. Both the ADALINE and the RLS algorithms give accurate results even under rapid dynamic changes. The paper investigates the effects of different parameters on the performance of the ADALINE algorithm and that of the RLS algorithm. Extensive simulations of the proposed flicker tracking algorithms are conducted to evaluate their performance for the tracking and mitigation of voltage flicker. The ADALINE and the RLS algorithms are examined by tracking and mitigating the flicker produced by a resistance welder and an arc furnace in a simple distribution system simulated in the PSCAD/EMTDC package.