Smoothing Factor Research Articles

Self-adaptive smoothing model for cardinality estimation

Abstract Cardinality estimation is a crucial component in query optimizers. After decades of research, employing autoregressive models for cardinality estimation has demonstrated remarkable accuracy. However, when queries involve attributes with large domain sizes, autoregressive model-based estimators struggle to accurately capture the data distribution, leading to poor performance. Furthermore, these models often exhibit significant errors when handling queries with low-selectivity predicates. To address these challenges, we propose a self-adaptive cardinality estimator named AdaCard. Initially, we employ a self-adaptive smoothing factor selection strategy to variably adjust the original data, thereby mitigating the impact of large domain sizes. Secondly, to correct errors stemming from Monte Carlo sampling, we utilize resampling to refine the handling of low-selectivity predicates, thereby improving accuracy. Through evaluation using four real-world benchmarks, we compared AdaCard with mainstream baselines. The final results show that our estimator has the lowest tail estimation error and improves accuracy by nearly 10$\times $ over the second-best method, with similar latency and model size.

Exact matrix completion via smooth matrix factorization

Exact matrix completion via smooth matrix factorization

Obstacle-aware path following of omni-wheeled robots using fuzzy inference approach

This study proposes a real-time LiDAR-based approach for efficient path following in omni-wheeled robots (OWRs) capable of dynamically adapting to surrounding obstacles. To achieve this, two fuzzy control schemes are developed: Fuzzy Sliding-Mode Tracking Control (FTC) for precise path following and Fuzzy Obstacle-avoidance Control (FOC) for real-time collision avoidance using LiDAR data. FTC utilizes fuzzified sliding surfaces, derived from tracking errors, to follow a planned path. In the presence of obstacles along the path, FOC, employing LiDAR measurements, assesses collision distance and direction, ensuring safe navigation. The outputs of FTC and FOC are seamlessly merged using a smooth switching factor to generate voltage inputs for the actuators. The effectiveness of the proposed approach is demonstrated through extensive simulations and real-world experiments under various conditions, including wheel disturbances, diverse trajectories, and varying initial positions.

RLI-SLAM: Fast Robust Ranging-LiDAR-Inertial Tightly-Coupled Localization and Mapping.

Simultaneous localization and mapping (SLAM) is an essential component for smart robot operations in unknown confined spaces such as indoors, tunnels and underground. This paper proposes a novel tightly-coupled ranging-LiDAR-inertial simultaneous localization and mapping framework, namely RLI-SLAM, which is designed to be high-accuracy, fast and robust in the long-term fast-motion scenario, and features two key innovations. The first one is tightly fusing the ultra-wideband (UWB) ranging and the inertial sensor to prevent the initial bias and long-term drift of the inertial sensor so that the point cloud distortion of the fast-moving LiDAR can be effectively compensated in real-time. This enables high-accuracy and robust state estimation in the long-term fast-motion scenario, even with a single ranging measurement. The second one is deploying an efficient loop closure detection module by using an incremental smoothing factor graph approach, which seamlessly integrates into the RLI-SLAM system, and enables high-precision mapping in a challenging environment. Extensive benchmark comparisons validate the superior accuracy of the proposed new state estimation and mapping framework over other state-of-the-art systems at a low computational complexity, even with a single ranging measurement and/or in a challenging environment.

Study on precise fuel injection under multiple injections of high pressure common rail system based on deep learning

The fluctuation of fuel injection quantity under multiple injections of high pressure common rail system leads to a decline in engine economy and emission performance, the precise injection control is the key problem in current research. In this study, a deep learning data-driven model of injection quantity under multiple injections was established based on generalized regression neural network (GRNN), and the particle swarm optimization (PSO) was introduced to optimize the smoothing factor in the model, thereby improving its prediction accuracy. An injection quantity correction method based on the data-driven model was proposed and applied to actual common rail fuel injection system for experimental verification. The results demonstrate that the proposed data-driven model has excellent prediction performance and generalization ability for the predicted main injection quantity with mean absolute percentage error of 1.10 % and determination coefficient of 0.997. Through the experimental verification of the correction method under 144 groups of operating conditions, it was observed that the maximum deviation of main injection quantity decreased from −25.83～17.91 mm3 to −9.81～7.64 mm3, and the maximum absolute deviation reduced by 62.02 %. The precise of injection control under multiple injections is improved.

Threshold approximations for the exponential of a factorized operator family with correctors taken into account

In a Hilbert space H \mathfrak H , consider a family of selfadjoint operators (a quadratic operator pencil) A ( t ) A(t) , t ∈ R t\in \mathbb {R} , of the form A ( t ) = X ( t ) ∗ X ( t ) A(t) = X(t)^* X(t) , where X ( t ) = X 0 + t X 1 X(t) = X_0 + t X_1 . It is assumed that the point λ 0 = 0 \lambda _0=0 is an isolated eigenvalue of finite multiplicity for the operator A ( 0 ) A(0) . Let F ( t ) F(t) be the spectral projection of the operator A ( t ) A(t) for the interval [ 0 , δ ] [0,\delta ] . Approximations for F ( t ) F(t) and A ( t ) F ( t ) A(t)F(t) for | t | ≤ t 0 |t| \leq t_0 (the so-called threshold approximations) are used to obtain approximations in the operator norm on H \mathfrak H for the operator exponential exp ⁡ ( − i τ A ( t ) ) \exp (-i \tau A(t)) , τ ∈ R \tau \in \mathbb {R} . The numbers δ \delta and t 0 t_0 are controlled explicitly. Next, the behavior for small ε > 0 \varepsilon >0 of the operator exp ⁡ ( − i ε − 2 τ A ( t ) ) \exp (-i \varepsilon ^{-2} \tau A(t)) multiplied by the “smoothing factor” ε s ( t 2 + ε 2 ) − s / 2 \varepsilon ^s (t^2 + \varepsilon ^2)^{-s/2} with a suitable s > 0 s>0 is studied. The obtained approximations are given in terms of the spectral characteristics of the operator A ( t ) A(t) near the lower edge of the spectrum. The results are aimed at application to homogenization of the Schrödinger-type equations with periodic rapidly oscillating coefficients.

Global Path Planning for Articulated Steering Tractor Based on Multi-Objective Hybrid Algorithm.

With the development of smart agriculture, autopilot technology is being used more and more widely in agriculture. Because most of the current global path planning only considers the shortest path, it is difficult to meet the articulated steering tractor operation needs in the orchard environment and address other issues, so this paper proposes a hybrid algorithm of an improved bidirectional search A* algorithm and improved differential evolution genetic algorithm(AGADE). First, the integrated priority function and search method of the traditional A* algorithm are improved by adding weight influence to the integrated priority, and the search method is changed to a bidirectional search. Second, the genetic algorithm fitness function and search strategy are improved; the fitness function is set as the path tree row center offset factor; the smoothing factor and safety coefficient are set; and the search strategy adopts differential evolution for cross mutation. Finally, the shortest path obtained by the improved bidirectional search A* algorithm is used as the initial population of an improved differential evolution genetic algorithm, optimized iteratively, and the optimal path is obtained by adding kinematic constraints through a cubic B-spline curve smoothing path. The convergence of the AGADE hybrid algorithm and GA algorithm on four different maps, path length, and trajectory curve are compared and analyzed through simulation tests. The convergence speed of the AGADE hybrid algorithm on four different complexity maps is improved by 92.8%, 64.5%, 50.0%, and 71.2% respectively. The path length is slightly increased compared with the GA algorithm, but the path trajectory curve is located in the center of the tree row, with fewer turns, and it meets the articulated steering tractor operation needs in the orchard environment, proving that the improved hybrid algorithm is effective.

Optimization of Fast Non-Local Means Noise Reduction Algorithm Parameter in Computed Tomographic Phantom Images Using 3D Printing Technology.

Noise in computed tomography (CT) is inevitably generated, which lowers the accuracy of disease diagnosis. The non-local means approach, a software technique for reducing noise, is widely used in medical imaging. In this study, we propose a noise reduction algorithm based on fast non-local means (FNLMs) and apply it to CT images of a phantom created using 3D printing technology. The self-produced phantom was manufactured using filaments with similar density to human brain tissues. To quantitatively evaluate image quality, the contrast-to-noise ratio (CNR), coefficient of variation (COV), and normalized noise power spectrum (NNPS) were calculated. The results demonstrate that the optimized smoothing factors of FNLMs are 0.08, 0.16, 0.22, 0.25, and 0.32 at 0.001, 0.005, 0.01, 0.05, and 0.1 of noise intensities, respectively. In addition, we compared the optimized FNLMs with noisy, local filters and total variation algorithms. As a result, FNLMs showed superior performance compared to various denoising techniques. Particularly, comparing the optimized FNLMs to the noisy images, the CNR improved by 6.53 to 16.34 times, COV improved by 6.55 to 18.28 times, and the NNPS improved by 10-2 mm2 on average. In conclusion, our approach shows significant potential in enhancing CT image quality with anthropomorphic phantoms, thus addressing the noise issue and improving diagnostic accuracy.

IMPLEMENTATION OF A FINGERPRINT ELECTRONIC ATTENDANCE POLICY AS A SUPPORT FOR EMPLOYEE WORK DISCIPLINE AT THE BUSINESS MANAGEMENT AGENCY (BPU) OF GORONTALO STATE UNIVERSITY

<em>This study aims to determine and describe (1) the implementation of the fingerprint electronic attendance policy as a support for employee work discipline at the Business Management Agency (BPU) of Gorontalo State University, which is studied from: 1. planning, 2. implementation, and 3. evaluation. (2) Determining Factors for the Success of the Implementation of the Finger Print Electronic Attendance Policy as a Support for Employee Work Discipline at the Business Management Agency (BPU) of Gorontalo State University, which are studied from: 1. Apparatus Compliance; 2. Smooth Absence of Problems; and 3. Satisfactory Performance. The approach used in this study is qualitative with descriptive qualitative methods. Data collection using interviews, observation, and documentation The results of this study show that (1) the implementation of the Fingerprint Electronic Attendance Policy (Fingerprint) as a Support for Employee Work Discipline at the Business Management Agency (BPU) of Gorontalo State University has run quite optimally. This is shown by the existence of a) planning, b) implementation, and c) evaluation that has been carried out based on the regulation of the Rector of Gorontalo State University Number 1,786 / UN47. A2/KP/2022 on July 25, 2022, to increase discipline, productivity, and work effectiveness, as well as working hours for civil servants in education and contract personnel within Gorontalo State University. Determining Factors for the Success of the Implementation of the Electronic Attendance Policy Fingerprint (Fingerprint) as a Support for Employee Work Discipline at the Business Management Agency (BPU) of Gorontalo State University based on the results of research conducted that has run optimally. This can be seen from a) apparatus compliance; employees at the Business Management Agency have shown compliance with regulations related to the use of electronic attendance to improve employee discipline, b) smoothness factors, the use of electronic attendance systems can increase the effectiveness in managing attendance administration as a guideline for employee benefits, c) satisfactory performance, employee performance, and productivity are increasing with the innovation of changes. The previous attendance method was still manual, despite technology</em>

Confidence-Aware Paced-Curriculum Learning by Label Smoothing for Surgical Scene Understanding

Curriculum learning and self-paced learning are the training strategies that gradually feed the samples from easy to more complex. They have captivated increasing attention due to their excellent performance in robotic vision. Most recent works focus on designing curricula based on difficulty levels in input samples or smoothing the feature maps. However, smoothing labels to control the learning utility in a curriculum manner is still unexplored. In this work, we design a paced curriculum by label smoothing (P-CBLS) using paced learning with uniform label smoothing (ULS) for classification tasks and fuse uniform and spatially varying label smoothing (SVLS) for semantic segmentation tasks in a curriculum manner. In ULS and SVLS, a bigger smoothing factor value enforces a heavy smoothing penalty in the true label and limits learning less information. Therefore, we design the curriculum by label smoothing (CBLS). We set a bigger smoothing value at the beginning of training and gradually decreased it to zero to control the model learning utility from lower to higher. We also designed a confidence-aware pacing function and combined it with our CBLS to investigate the benefits of various curricula. The proposed techniques are validated on four robotic surgery datasets of multi-class, multi-label classification, captioning, and segmentation tasks. We also investigate the robustness of our method by corrupting validation data into different severity levels. Our extensive analysis shows that the proposed method improves prediction accuracy and robustness. The code is publicly available at https://github.com/XuMengyaAmy/P-CBLS. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The motivation of this article is to improve the performance and robustness of deep neural networks in safety-critical applications such as robotic surgery by controlling the learning ability of the model in a curriculum learning manner and allowing the model to imitate the cognitive process of humans and animals. The designed approaches do not add parameters that require additional computational resources.

An Adaptive Fast Incremental Smoothing Approach to INS/GPS/VO Factor Graph Inference

In response to asynchronous and delayed sensors within multi-sensor integrated navigation systems, the computational complexity of joint optimization navigation solutions persistently rises. This paper introduces an adaptive fast integrated navigation algorithm for INS/GPS/VO based on factor graph. The factor graph model for INS/GPS/VO is developed subsequent to individual modeling of the Inertial Navigation System (INS), Global Positioning System (GPS), and Visual Odometer (VO) using the factor graph model approach. Additionally, an Adaptive Fast Incremental Smoothing (AFIS) factor graph optimization algorithm is proposed. The simulation results demonstrate that the factor-graph-based integrated navigation algorithm consistently yields high-precision navigation outcomes even amidst dynamic changes in sensor validity and the presence of asynchronous and delayed sensor measurements. Notably, the AFIS factor graph optimization algorithm significantly enhances real-time performance compared to traditional Incremental Smoothing (IF) algorithms, while maintaining comparable real-time accuracy.

On the design of Optimal Health Indicators for early fault detection and their statistical thresholds

This research presents a method for developing optimal health indicators for early fault detection in machine condition monitoring. The devised health indicators are optimal in the Neyman–Pearson sense, as they maximize the probability of detection while maintaining a constant false alarm rate. They are generated through a generalized likelihood ratio test that involves the probability density functions representing the healthy and faulty states. The essence of the approach lies in establishing a framework that defines optimal health indicators as functions of the partial derivatives of the healthy probability density function and a modulation function, representing the distortion of the probability density function in the faulty state. Specific modulation functions lead to the recovery of several traditional health indicators, including kurtosis, skewness, the crest factor, the clearance factor, the smoothness factor, the shape factor, the impulse factor, ℓp/ℓq-norms, the negentropy, among others. New health indicators are also derived. As the proposed optimal health indicators asymptotically follow a chi-squared distribution, they come with a statistical threshold. The study comprehensively outlines the methodology, with its performance assessed through simulated and experimental signals. The results reveal that the early transition from healthy to faulty can be well detected via the established optimal health indicators and their thresholds.

Automated thorax disease diagnosis using multi-branch residual attention network

Chest X-ray (CXR) is an extensively utilized radiological modality for supporting the diagnosis of chest diseases. However, existing research approaches suffer from limitations in effectively integrating multi-scale CXR image features and are also hindered by imbalanced datasets. Therefore, there is a pressing need for further advancement in computer-aided diagnosis (CAD) of thoracic diseases. To tackle these challenges, we propose a multi-branch residual attention network (MBRANet) for thoracic disease diagnosis. MBRANet comprises three components. Firstly, to address the issue of inadequate extraction of spatial and positional information by the convolutional layer, a novel residual structure incorporating a coordinate attention (CA) module is proposed to extract features at multiple scales. Next, based on the concept of a Feature Pyramid Network (FPN), we perform multi-scale feature fusion in the following manner. Thirdly, we propose a novel Multi-Branch Feature Classifier (MFC) approach, which leverages the class-specific residual attention (CSRA) module for classification instead of relying solely on the fully connected layer. In addition, the designed BCEWithLabelSmoothing loss function improves the generalization ability and mitigates the problem of class imbalance by introducing a smoothing factor. We evaluated MBRANet on the ChestX-Ray14, CheXpert, MIMIC-CXR, and IU X-Ray datasets and achieved average AUCs of 0.841, 0.895, 0.805, and 0.745, respectively. Our method outperformed state-of-the-art baselines on these benchmark datasets.

Application and characterization of digital Sallen–Key filter in nuclear spectrum smoothing

In the nuclear spectrum analysis processing, spectrum smoothing can remove the statistical fluctuation in the spectrum, which is beneficial for peak detection and peak area calculation. In this work, a spectrum smoothing algorithm is proposed based on digital Sallen–Key filter, which contains four parameters (m, n, k, D). The amplitude–frequency response curve of Sallen–Key filter is deduced and the filtering performance is analyzed. Meanwhile, the effects of the four parameters on the shape of the smoothed spectrum are explored: D affects the counts and peak areas of the spectrum, and the peak area can be corrected by the peak area correction function S’. The parameters of m, n and k affect the peak position after smoothing, making the peak position shift to the right, and the peak position correction function P’ can be used to correct the peak position, when n¿2, the spectrum data appear negative after smoothing, when k¿2, the smoothed spectrum broadening degree is greater than 20%. Smoothness (R), noise smoothing factor (NSF), spectrum count ratio before and after smoothing (PER), and comprehensive evaluation factor (Q) are used to evaluate the smoothing effect of the algorithm. The parameters of the algorithm are optimally selected: about the gamma spectrum of 137Cs and 60Co, the optimal parameters are m=1.5 n=2 k=2 D=1, about the characteristic X-ray spectrum of Fe and quasi-geological sample (TiMnFeNiCuZn), the optimal parameters are m=1.1 n=1.1 k=1.3 D=1. Based on Sallen–Key smoothing method, Fourier transform method, Gaussian function method, wavelet transformation method, center of gravity method and least squares method, the gamma spectrum of 137Cs is smoothed and denoised in this paper. The results show that the Sallen–Key method has better spectrum denoising effect (R=0.6056) and comprehensive performance indicators (Q=0.6104), which can be further applied for the smoothing of nuclear spectrum data.

Vascular smooth muscle cell-derived nerve growth factor regulates sympathetic collateral branching to innervate blood vessels in embryonic skin.

Blood vessels serve as intermediate conduits for the extension of sympathetic axons towards target tissues, while also acting as crucial targets for their homeostatic processes encompassing the regulation of temperature, blood pressure, and oxygen availability. How sympathetic axons innervate not only blood vessels but also a wide array of target tissues is not clear. Here we show that in embryonic skin, after the establishment of co-branching between sensory nerves and blood vessels, sympathetic axons invade the skin alongside these sensory nerves and extend their branches towards these blood vessels covered by vascular smooth muscle cells (VSMCs). Our mosaic labeling technique for sympathetic axons shows that collateral branching predominantly mediates the innervation of VSMC-covered blood vessels by sympathetic axons. The expression of nerve growth factor (NGF), previously known to induce collateral axon branching in culture, can be detected in the vascular smooth muscle cell (VSMC)-covered blood vessels, as well as sensory nerves. Indeed, VSMC-specific Ngf knockout leads to a significant decrease of collateral branching of sympathetic axons innervating VSMC-covered blood vessels. These data suggest that VSMC-derived NGF serves as an inductive signal for collateral branching of sympathetic axons innervating blood vessels in the embryonic skin.

Baseline estimation using optimized asymmetric least squares (O-ALS)

Baseline drift widely exists in the spectral measurements due to instrumental nonlinearity. It negatively affects the qualitative or quantitative analysis of chemical and physical parameters, thus, an accurate baseline estimation is needed. An improved algorithm, optimized asymmetric least squares smoothing (O-ALS), is designed based on the traditional ALS algorithm to achieve effective estimation. First, an appropriate smooth factor is selected according to the number of second-order concave points (NSCP) of a measured spectrum. Then, the asymmetric penalty factor is adaptively optimized in each iteration based on the differences between the estimated baseline and the measured spectrum. Various simulated and real measured data, including Fourier Transform Spectrometer spectrum, chromatogram, and ion mobility spectroscopy, have been used for evaluation. All results achieve a high precision, which suggests that our algorithm is effective, flexible, and robust.

Forecasting regional water demand using multi-fidelity data and harris hawks optimization of generalized regression neural network models – A case study of Heilongjiang Province, China

Accurate forecasting of water demand plays a crucial role in decision makers to effectively allocate regional water resources and promotes the rational development and utilization of water resources. To address the challenges posed by limited raw data availability and the difficulty in selecting parameters for water demand forecasting models, this study proposes a coupled model called MF-HHO-GRNN (Multi-Fidelity Data, Hybrid Harris Hawks Optimization, and Generalized Regression Neural Network). Initially, multi-fidelity data was constructed employing water demand factor data for Heilongjiang Province, China (1999–2020) as high-fidelity data, alongside data from Liaoning and Jilin Provinces of China as low-fidelity data. Adaptive boosting algorithms, data concatenation, and data dimensionality reduction techniques were employed to construct multi-fidelity data, thereby improving the availability of the original data. Subsequently, Harris hawk optimization algorithms were implemented to optimize the smoothing factor of the GRNN to resolve the challenge of selecting suitable parameters for water demand forecasting models. Finally, the MF-HHO-GRNN model was compared with other prediction models to analyze the prediction accuracy. The optimized MF-HHO-GRNN model meeting accuracy requirements was utilized to forecast agricultural irrigation, industrial, domestic, ecological water demand, and total water demand in Heilongjiang Province for the next 5 years. Results demonstrated that the proposed MF-HHO-GRNN model attained predictive accuracy of 98 %, 98 %, and 97 % for total water demand under three distinct time horizons (1 year, 3 years, and 5 years), proving excellent predictive capabilities. Furthermore, a comprehensive comparison was made between the predicted total water demand and user-specific water demands for the next 5 years in Heilongjiang Province and the regional development plans. The results showed that the predictions of the coupled model aligned with the future development trends of Heilongjiang Province, validating the practicality of the MF-HHO-GRNN model. This study provides decision-makers with an effective method for forecasting water demand, contributing to the future realization of more intelligent and accurate water resource management and decision-making.

Quantitative analysis of fast non-local means algorithm smoothing factor impact on lung segmentation accuracy in computed tomography images

In computed tomography (CT), noise inevitably occurs during image acquisition and degrades segmentation accuracy. Thus, in this study, we compared and evaluated quantitative factors to derive the optimized smoothing factor of the fast non-local means (FNLM) algorithm, which can improve the accuracy of lung segmentation. Gaussian noise with a standard deviation of 0.05 was added to the lung phantom image and clinically acquired lung image, respectively, and the smoothing factor of the FNLM algorithm was modeled and applied from 0.03 to 0.02 in 0.01 intervals. The segmentation algorithm is modeled based on a region-growing approach. For evaluating segmentation performance, the root mean square error (RMSE), peak signal-to-noise ratio (PSNR) and dice score (DSC) were used. The quantitative evaluation demonstrated that the RMSE, PSNR and DSC exhibited the greatest improvement when the smoothing factor was set as 0.10 for both the phantom and clinically acquired lung images. In the lung phantom images, the RMSE, PSNR, and DSC displayed improvements of 8.40, 1.33, and 1.33 times compared to the noisy images. In the clinically acquired lung images, the RMSE, PSNR, and DSC exhibited enhancements of 9.74, 1.35, 1.37 times respectively, compared to the noisy images. We confirmed that lung segmentation performance was most effectively improved when the smoothing factor of the FNLM algorithm was set to 0.10. Moreover, we demonstrated that adjusting the appropriate smoothing factor in CT images can improve lung segmentation performance.

Data-driven smoothing approaches for interest modeling in recommendation systems

In recommendation systems, users often click on some items that are distinct from historically clicked items. This verifies the existence of interest gaps between the historical interests reflected by historical behaviors and current interests. The gaps pose key challenges for interest modeling in recommendation systems, i.e., learn user current interests from historical behaviors. To bridge interest gaps, various solutions have been proposed from a model perspective, and mainly focus on capturing interest diversity and interest evolution process (the changes in user interests over time). Different from existing solutions, our key idea is smoothing the historical interests from a data perspective, so that interest gaps can be smoothed and bridged. To realize this idea, we propose two types of data-driven smoothing approaches based on the principle of collaborative filtering, i.e., click-centric smoothing and user-centric smoothing. The approaches can be easily applied to a number of existing solutions and enhance their performances. To ensure that the smoothed interests completely and accurately cover user current interests (coverage and accuracy), the click-centric smoothing leverages the relevant clicks of a click to smooth the interests on the click, and the user-centric smoothing leverages the relevant clicks of a user click sequence to smooth the interests learned from the click sequence. The “smooth” is to enhance the coverage, and “relevant” is to enhance the accuracy. To search the relevant clicks, we develop an iterative Bayesian inference over a click-flow graph, i.e., a new implementation of collaborative filtering. To further enhance the accuracy, we propose new constrained attention networks to generate high-quality smoothing factors from these relevant clicks. Both offline and online experiments show the data-driven smoothing approaches significantly outperform the baselines. The user-centric smoothing has been deployed on the e-commerce platform of Meituan and achieved 1.51% elevation of Click-through & Conversion Rate and 0.53% elevation of Click-through Rate.

Transformers in High-Frequency Trading

Transformer is a deep learning model that, having an innovative performance in many tasks, has uniquely and significantly modified all the cast of mind of the AI scientific community. In this paper, we introduce a Transformer model that is applied to 1-minute timescale in the EURUSD and GBPUSD instruments of forex trading. We use the classical Transformer architecture without the Decoder since the Encoder is enough. Moreover, the Exponential Moving Average (EMA) model is applied to the input, while different values of its smoothing factor α are tested. With cross-entropy training loss less than 0.2, it is exclamatory that Transformers are a promising tool for lucrative strategies in high-frequency trading.