This study presents a facial emotion recognition network based on UniRepLKNet to address the difficulty in effectively capturing feature information and preventing key facial information from occupying a more prominent position in the facial emotion recognition process. Moreover, to extract facial emotional features more accurately, the study designs a masked polarized self-attention module that combines U-Net and a polarized self-attention mechanism. This module can deeply mine the dependency between channels and spaces. It can also strengthen the influence of local key information of the face on emotion recognition through a multi-scale feature fusion strategy. The study optimizes UniRepLKNet, a universal large kernel Convolutional Neural Network (CNN), and proposes the EmoRepLKNet neural network structure. In EmoRepLKNet, the mask-polarized self-attention module enables the network to extract key information for facial emotion recognition. Combined with the wide receptive field of large kernel CNN, facial emotions can be recognized effectively. Experimental results show that on the facial emotion recognition dataset FER2013, EmoRepLKNet achieves an accuracy of 76.20%, outperforming existing comparison models and significantly improving facial emotion recognition accuracy compared to that of UniRepLKNet. Additionally, on the single-label portion of the RAF-DB dataset, the proposed method achieves an accuracy of 89.67%.
Abstract Matrix geometric means between two positive definite matrices can be defined from distinct perspectives—as solutions to certain nonlinear systems of equations, as points along geodesics in Riemannian geometry, and as solutions to certain optimisation problems. We devise quantum subroutines for the matrix geometric means, and construct solutions to the algebraic Riccati equation—an important class of nonlinear systems of equations appearing in machine learning, optimal control, estimation, and filtering. Using these subroutines, we present a new class of quantum learning algorithms, for both classical and quantum data, called quantum geometric mean metric learning, for weakly supervised learning and anomaly detection. The subroutines are also useful for estimating geometric Rényi relative entropies and the Uhlmann fidelity, in particular achieving optimal dependence on precision for the Uhlmann and Matsumoto fidelities. Finally, we provide a BQP-complete problem based on matrix geometric means that can be solved by our subroutines.
Emmanuel Gottlob, Dan S. Borgnia, Robert-Jan Slager
et al.
The robustness of topological properties, such as quantized currents, generally depends on the existence of gaps surrounding the relevant energy levels or on symmetry-forbidden transitions. Here, we observe quantized currents that survive the addition of bounded local disorder beyond the closing of the relevant instantaneous energy gaps in a driven Aubry-André-Harper chain, a prototypical model of quasiperiodic systems. We explain the robustness using a local picture in configuration space based on Landau-Zener transitions, which rests on the Anderson localization of the eigenstates. Moreover, we propose a protocol, directly realizable in, for instance, cold atoms or photonic experiments, that leverages this stability to prepare topological many-body states with high Chern numbers and opens new experimental avenues for the study of both the integer and fractional quantum Hall effects.
Abstract To leverage advancements in machine learning for metallic materials design and property prediction, it is crucial to develop a data-reduced representation of metal microstructures that surpasses the limitations of current physics-based discrete microstructure descriptors. This need is particularly relevant for metallic materials processed through additive manufacturing, which exhibit complex hierarchical microstructures that cannot be adequately described using the conventional metrics typically applied to wrought materials. Furthermore, capturing the spatial heterogeneity of microstructures at the different scales is necessary within such framework to accurately predict their properties. To address these challenges, we propose the physical spatial mapping of metal diffraction latent space features. This approach integrates (i) point diffraction data encoding via variational autoencoders or contrastive learning and (ii) the physical mapping of the encoded values. Together, these steps offer a method to comprehensively describe metal microstructures. We demonstrate this approach on a wrought and additively manufactured alloy, showing that it effectively encodes microstructural information and enables direct identification of microstructural heterogeneity not directly possible by physics-based models. This data-reduced microstructure representation opens the application of machine learning models in accelerating metallic material design and accurately predicting their properties.
Materials of engineering and construction. Mechanics of materials, Computer software
Hassan-Roland Nasser, Marianne Cockburn, Marie Schneider
Studying animals’ rhythmicity provides insights into their physiological and psychological states. The degree of functional coupling (DFC) is one of the algorithms available to assess rhythmicity in activity-related time series data, such as accelerometer or GPS data. However, DFC computation is complex, as it includes frequency spectrum analysis and statistical significance testing. This paper introduces digiRhythm, an R package that makes the DFC-based rhythmicity analysis easily accessible. Beyond the DFC, the package includes an additional set of tools, which are crucial for rhythmicity investigations, such as actogram generation, daily activity visualization, and diurnality index computation.
Scientific data acquisition is a problem domain that has been underserved by its computational tools despite the need to efficiently use hardware, to guarantee validity of the recorded data, and to rapidly test ideas by configuring experiments quickly and inexpensively. High-dimensional physical spectroscopies, such as angle-resolved photoemission spectroscopy, make these issues especially apparent because, while they use expensive instruments to record large data volumes, they require very little acquisition planning. The burden of writing data acquisition software falls to scientists, who are not typically trained to write maintainable software. In this paper, we introduce AutodiDAQt to address these shortfalls in the scientific ecosystem. To ground the discussion, we demonstrate its merits for angle-resolved photoemission spectroscopy and high bandwidth spectroscopies. AutodiDAQt addresses the essential needs for scientific data acquisition by providing simple concurrency, reproducibility, retrospection of the acquisition sequence, and automated user interface generation. Finally, we discuss how AutodiDAQt enables a future of highly efficient machine-learning-in-the-loop experiments and analysis-driven experiments without requiring data acquisition domain expertise by using analysis code for external data acquisition planning.
Human-computer interaction and computer-mediated behavioral psychology research studies often rely on capturing user interaction data to characterize online behaviors. IRB considerations, site policies, and/or security and privacy concerns may force researchers to use screenshots or offline copies of pages of interest, instead of live websites, in their study designs. These interaction modalities reduce the fidelity and contextual realism of web content and often affect interface aesthetic quality – due to broken links, missing images, and/or malfunctioning scripts. StudySandboxx is a tool that allows websites to be saved exactly as they appear online. The tool sandboxes websites in a way that removes dangerous scripts that threaten privacy and security. Saved websites are encapsulated into a single portable file that contains all related website resources. Finally, the tool also supports certain types of permutations commonly used in research – such as changing links in a page. The project is housed within a GitHub repository at https://github.com/gewethor/study-sandbox.
Abstract With the continuous and rapid growth of transport demand, scheduling strategy of warehouse has become a key issue in the field of logistics transportation. The structural differences of the warehouse, the automated storage and retrieval system (AS/RS) model and the two‐end dual stackers scheduling model (TDSM) are considered, and a new improved genetic algorithm (NIGA) is proposed. It can adjust the algorithm structure according to the density of population fitness value, and effectively optimize the stacker path. In the TDSM, an improved anti‐collision principle is proposed to avoid collision of two stackers. Besides, combined with the optimal anti‐collision boundary inspection mechanism, the best working area for the two stackers is allocated by using NIGA. Finally, the new improved GA is compared with GA and the adaptive GA on specific storage and retrieval tasks. The simulation results show that the proposed NIGA well outperforms other GAs in most instances, which indicates that it is an effective approach for the AS/RS and the TDSM scheduling optimization problem.
Computational linguistics. Natural language processing, Computer software
A Vehicular Ad-Hoc Network(VANET) is a Mobile Ad-Hoc Network(MANET) composed of mobile vehicular nodes.It does not rely on infrastructure to either establish a communication link orrealize communication. Owingto the high mobility of vehicles and limited wireless-communication resources, it is difficult for VANETs to guarantee Quality of Service(QoS).To solve this problem, this paper introduces a Software-Defined Network(SDN).In particular, amulti-constrained QoS routing algorithm suitable for Doftware-Defined Vehicular Ad-Hoc Network(SDN-VANET) is proposedthatharnessesthe advantages of SDN control and forwarding separation to ensurevehicle QoS.First, the SDN controller schedules a vehicle's service based on deadline constraints.Second, this paper proposesan Adaptive Hybrid Shuffled Frog-Leaping Algorithm(AH-SFLA).The SDN controller calculates the appropriate value of the data on the transmission link according to the QoS index and the global topology information and uses this as a benchmark to search for an optimized path.At the same time, alternative link mechanisms and QoS resource consumption thresholds are set to implement routing maintenance in order toreduce the probability of network failures.Finally, mininet-wifi and SUMO are combined to build an SDN-VANET environment, and the AH-SFLA routing algorithm is compared with the performances ofIGA and IICSFLA.The experimental results show that compared with IGA and IICSFL, AH-SFLA can improve the average end-to-end delay index by 57.74% and 46.6%, reduce the packet-loss rate by 29.9% and 18.6%, and increase the cost of standardized routing by 36.93% and 27.2%, respectively, effectively guaranteeingQoS in VANET.
Nowadays, robots are found in a growing number of areas where they collaborate closely with humans. Enabled by lightweight materials and safety sensors, these cobots are gaining increasing popularity in domestic care, where they support people with physical impairments in their everyday lives. However, when cobots perform actions autonomously, it remains challenging for human collaborators to understand and predict their behavior, which is crucial for achieving trust and user acceptance. One significant aspect of predicting cobot behavior is understanding their perception and comprehending how they “see” the world. To tackle this challenge, we compared three different visualization techniques for Spatial Augmented Reality. All of these communicate cobot perception by visually indicating which objects in the cobot’s surrounding have been identified by their sensors. We compared the well-established visualizations <i>Wedge</i> and <i>Halo</i> against our proposed visualization <i>Line</i> in a remote user experiment with participants suffering from physical impairments. In a second remote experiment, we validated these findings with a broader non-specific user base. Our findings show that <i>Line</i>, a lower complexity visualization, results in significantly faster reaction times compared to <i>Halo</i>, and lower task load compared to both <i>Wedge</i> and <i>Halo</i>. Overall, users prefer <i>Line</i> as a more straightforward visualization. In Spatial Augmented Reality, with its known disadvantage of limited projection area size, established off-screen visualizations are not effective in communicating cobot perception and <i>Line</i> presents an easy-to-understand alternative.
XU Le, AN Hong, CHEN Junshi, ZHANG Pengfei, WU Zheng
The performance of unstructured grid computing on Sunway TaihuLight, a domestic heterogeneous many-core platform, is limited by sparse storage, discrete memory access, and data dependency.To relieve the sparse storage and discrete memory access problems, this paper proposes an N-order diagonal coloring algorithm, which effectively balances the computing between Management Processing Element (MPE) and Computing Processing Elements (CPEs) and convert global memory access to Local Device Memory (LDM) access using CPEs.To solve the computing competition caused by data dependence, this paper presents an adaptive and independent blocking method to avoid data conflicts in parallel computing.Furthermore, various optimizations are employed to overcome the performance bottlenecks:1.To leverage hardware resources, the authors use asynchronous parallelism between MPE and CPEs.2.To reduce synchronization costs, they avoid register communication, which increases the scalability of the next-generation Sunway platform.3.To hide the memory access latency, the authors overlap memory access with computing.The SpMV, Integration, and calcLudsFcc operations are generally used to verify the validity of the algorithm, and the results show that our algorithm achieves an average speedup of about 10 times and up to 24 times higher than that of the MPE implementation.Moreover, the N-order diagonal coloring algorithm has a 5.8 times higher speedup than that of the non-coloring blocking algorithm, which effectively improves data locality and computational parallelism.The algorithm also has good acceleration performance for dependent conflict operators, which verifies the effectiveness of adaptive and independent task partitioning methods.
Different from artistic style transfer,the challenge of photorealistic style transfer is to maintain the authenticity of the output while transferring the color style of the style input.Now,most photorealistic style transfer methods perform pre-proces-sing or post-processing based on artistic style transfer methods,to maintain the authenticity of the output image.However,artistic style transfer methods usually cannot make full use of global color information to achieve a more coordinated overall impression,and pre-processing and post-processing operations are often tedious and time-consuming.To solve the above problems,this paper establishes a photorealistic style transfer network guided by global information,and proposes a color-partition-mean loss(<i>L<sub>cpm</sub></i>) to measure the similarity of the global color distribution between output and the style input.Adaptive instance normalization(AdaIN) is improved,and partition adaptive instance normalization(AdaIN-P) is proposed to better adapt to the color style transfer of real images.In addition,this paper also introduces a cross-channel partition attention module to make better use of global context information and improve the overall coordination of output images.Through the above methods,the decoder of network is guided to make full use of global information.Experimental results show that,compared with other state-of-the-art me-thods,the proposed model can achieve a better photorealistic style transfer effect while maintaining image details.
Pedro Teixidó, Juan Antonio Gómez-Galán, Rafael Caballero
et al.
Perimeter detection systems detect intruders penetrating protected areas, but modern solutions require the combination of smart detectors, information networks and controlling software to reduce false alarms and extend detection range. The current solutions available to secure a perimeter (infrared and motion sensors, fiber optics, cameras, radar, among others) have several problems, such as sensitivity to weather conditions or the high failure alarm rate that forces the need for human supervision. The system exposed in this paper overcomes these problems by combining a perimeter security system based on CEMF (control of electromagnetic fields) sensing technology, a set of video cameras that remain powered off except when an event has been detected. An autonomous drone is also informed where the event has been initially detected. Then, it flies through computer vision to follow the intruder for as long as they remain within the perimeter. This paper covers a detailed view of how all three components cooperate in harmony to protect a perimeter effectively, without having to worry about false alarms, blinding due to weather conditions, clearance areas, or privacy issues. The system also provides extra information of where the intruder is or has been, at all times, no matter whether they have become mixed up with more people or not during the attack.
Cache memory plays an important role in improving the performance of web servers, especially for big data transmission, which response time is constrained. It is necessary to use an effective method, such as web cache. Because an outstanding cache replacement algorithm can not only reduce the users access time but also improve the performance of the system. The traditional used weighting replacement policy does not consider the size parameter, hence, it may perform poorly while the datasets are larger. In this paper, we propose a novel, high-performance cache replacement algorithm for the web cache, named weighting size and cost replacement policy (WSCRP) bases on the weighting replacement policy. The algorithm recalculates the objects weight with adding the cost attribute in the cache, then orders the weight. Additionally the influence of various factors on the Web object as frequency, time, and cost value are considered. When the cache space cannot satisfy the new request object, the replacement policy WSCRP replaces the largest weighting and cost object. The experiments show that proposed algorithm has higher
Aiming at the theme drifting and the page weight splitting of traditional PageRank algorithm,an improved PageRank algorithm is proposed.In order to improve the user query efficiency and search quality,combined with the time feedback factor,it makes a comprehensive analysis on user forwarding,user comments and micro-blog mentions.Statistical analysis is used to measure the contribution of user behavior in the ranking of micro-blog user influence.By using the improved TF-IDF algorithm to calculate the similarity weight of the topic,the user can select the Web page with high relevance to obtain the corresponding PageRank weight.Experimental results show that compared with common microblog ranking algorithms,the improved PageRank algorithm has better user influence ranking effect.
In view of the defect existed in the time synchronization of Local Area Network(LAN) using hardware and software methods,this paper introduces the principle of time synchronization in LAN and the method of robust estimation.The model of time comparison is established.A new method for time synchronization in LAN using the robust estimation method is proposed,and uses the clock retaining strategy to save the overhead of CPU.The least squares adjustment method and the robust estimation method are respectively used for measuring data processing to analyze the accuracy difference.Experimental results show that the synchronization method has the advantages of simple hardware configuration with high synchronization precision,and it effectively restrains the influence of network delay while keeping the good stability of system.
The study of brain disorders requires accurate tissue segmentation of magnetic resonance (MR) brain images which is very important for detecting tumors, edema, and necrotic tissues. Segmentation of brain images, especially into three main tissue types: Cerebrospinal Fluid (CSF), Gray Matter (GM), and White Matter (WM), has important role in computer aided neurosurgery and diagnosis. Brain images mostly contain noise, intensity inhomogeneity, and weak boundaries. Therefore, accurate segmentation of brain images is still a challenging area of research. This paper presents a review of fuzzy c-means (FCM) clustering algorithms for the segmentation of brain MR images. The review covers the detailed analysis of FCM based algorithms with intensity inhomogeneity correction and noise robustness. Different methods for the modification of standard fuzzy objective function with updating of membership and cluster centroid are also discussed.