A novel gappy technology, gappy autoencoder with proper orthogonal decomposition(Gappy POD-AE), is proposed for reconstructing physical fields from sparse data. High-dimensional data are reduced via proper orthogonal ...A novel gappy technology, gappy autoencoder with proper orthogonal decomposition(Gappy POD-AE), is proposed for reconstructing physical fields from sparse data. High-dimensional data are reduced via proper orthogonal decomposition(POD),and low-dimensional data are used to train an autoencoder(AE). By integrating the POD operator with the decoder, a nonlinear solution form is established and incorporated into a new maximum-a-posteriori(MAP)-based objective for online reconstruction.The numerical results on the two-dimensional(2D) Bhatnagar-Gross-Krook-Boltzmann(BGK-Boltzmann) equation, wave equation, shallow-water equation, and satellite data show that Gappy POD-AE achieves higher accuracy than gappy proper orthogonal decomposition(Gappy POD), especially for the data with slowly decaying singular values,and is more efficient in training than gappy autoencoder(Gappy AE). The MAP-based formulation and new gappy procedure further enhance the reconstruction accuracy.展开更多
The proliferation of internet communication channels has increased telecom fraud,causing billions of euros in losses for customers and the industry each year.Fraudsters constantly find new ways to engage in illegal ac...The proliferation of internet communication channels has increased telecom fraud,causing billions of euros in losses for customers and the industry each year.Fraudsters constantly find new ways to engage in illegal activity on the network.To reduce these losses,a new fraud detection approach is required.Telecom fraud detection involves identifying a small number of fraudulent calls from a vast amount of call traffic.Developing an effective strategy to combat fraud has become challenging.Although much effort has been made to detect fraud,most existing methods are designed for batch processing,not real-time detection.To solve this problem,we propose an online fraud detection model using a Neural Factorization Autoencoder(NFA),which analyzes customer calling patterns to detect fraudulent calls.The model employs Neural Factorization Machines(NFM)and an Autoencoder(AE)to model calling patterns and a memory module to adapt to changing customer behaviour.We evaluate our approach on a large dataset of real-world call detail records and compare it with several state-of-the-art methods.Our results show that our approach outperforms the baselines,with an AUC of 91.06%,a TPR of 91.89%,an FPR of 14.76%,and an F1-score of 95.45%.These results demonstrate the effectiveness of our approach in detecting fraud in real-time and suggest that it can be a valuable tool for preventing fraud in telecommunications networks.展开更多
文章提出一种在片上系统(System on Chip,SoC)实现高吞吐率的有限状态熵编码(finite state entropy,FSE)算法。通过压缩率、速度、资源消耗、功耗4个方面对所提出的编码器和解码器与典型的硬件哈夫曼编码(Huffman coding,HC)进行性能比...文章提出一种在片上系统(System on Chip,SoC)实现高吞吐率的有限状态熵编码(finite state entropy,FSE)算法。通过压缩率、速度、资源消耗、功耗4个方面对所提出的编码器和解码器与典型的硬件哈夫曼编码(Huffman coding,HC)进行性能比较,结果表明,所提出的硬件FSE编码器和解码器具有显著优势。硬件FSE(hFSE)架构实现在SoC的处理系统和可编程逻辑块(programmable logic,PL)上,通过高级可扩展接口(Advanced eXtensible Interface 4,AXI4)总线连接SoC的处理系统和可编程逻辑块。算法测试显示,FSE算法在非均匀数据分布和大数据量情况下,具有更好的压缩率。该文设计的编码器和解码器已在可编程逻辑块上实现,其中包括1个可配置的缓冲模块,将比特流作为单字节或双字节配置输出到8 bit位宽4096深度或16 bit位宽2048深度的块随机访问存储器(block random access memory,BRAM)中。所提出的FSE硬件架构为实时压缩应用提供了高吞吐率、低功耗和低资源消耗的硬件实现。展开更多
Much like humans focus solely on object movement to understand actions,directing a deep learning model’s attention to the core contexts within videos is crucial for improving video comprehension.In the recent study,V...Much like humans focus solely on object movement to understand actions,directing a deep learning model’s attention to the core contexts within videos is crucial for improving video comprehension.In the recent study,Video Masked Auto-Encoder(VideoMAE)employs a pre-training approach with a high ratio of tube masking and reconstruction,effectively mitigating spatial bias due to temporal redundancy in full video frames.This steers the model’s focus toward detailed temporal contexts.However,as the VideoMAE still relies on full video frames during the action recognition stage,it may exhibit a progressive shift in attention towards spatial contexts,deteriorating its ability to capture the main spatio-temporal contexts.To address this issue,we propose an attention-directing module named Transformer Encoder Attention Module(TEAM).This proposed module effectively directs the model’s attention to the core characteristics within each video,inherently mitigating spatial bias.The TEAM first figures out the core features among the overall extracted features from each video.After that,it discerns the specific parts of the video where those features are located,encouraging the model to focus more on these informative parts.Consequently,during the action recognition stage,the proposed TEAM effectively shifts the VideoMAE’s attention from spatial contexts towards the core spatio-temporal contexts.This attention-shift manner alleviates the spatial bias in the model and simultaneously enhances its ability to capture precise video contexts.We conduct extensive experiments to explore the optimal configuration that enables the TEAM to fulfill its intended design purpose and facilitates its seamless integration with the VideoMAE framework.The integrated model,i.e.,VideoMAE+TEAM,outperforms the existing VideoMAE by a significant margin on Something-Something-V2(71.3%vs.70.3%).Moreover,the qualitative comparisons demonstrate that the TEAM encourages the model to disregard insignificant features and focus more on the essential video features,capturing more detailed spatio-temporal contexts within the video.展开更多
基金supported by the National Natural Science Foundation of China(No.12472197)。
文摘A novel gappy technology, gappy autoencoder with proper orthogonal decomposition(Gappy POD-AE), is proposed for reconstructing physical fields from sparse data. High-dimensional data are reduced via proper orthogonal decomposition(POD),and low-dimensional data are used to train an autoencoder(AE). By integrating the POD operator with the decoder, a nonlinear solution form is established and incorporated into a new maximum-a-posteriori(MAP)-based objective for online reconstruction.The numerical results on the two-dimensional(2D) Bhatnagar-Gross-Krook-Boltzmann(BGK-Boltzmann) equation, wave equation, shallow-water equation, and satellite data show that Gappy POD-AE achieves higher accuracy than gappy proper orthogonal decomposition(Gappy POD), especially for the data with slowly decaying singular values,and is more efficient in training than gappy autoencoder(Gappy AE). The MAP-based formulation and new gappy procedure further enhance the reconstruction accuracy.
基金This research work has been conducted in cooperation with members of DETSI project supported by BPI France and Pays de Loire and Auvergne Rhone Alpes.
文摘The proliferation of internet communication channels has increased telecom fraud,causing billions of euros in losses for customers and the industry each year.Fraudsters constantly find new ways to engage in illegal activity on the network.To reduce these losses,a new fraud detection approach is required.Telecom fraud detection involves identifying a small number of fraudulent calls from a vast amount of call traffic.Developing an effective strategy to combat fraud has become challenging.Although much effort has been made to detect fraud,most existing methods are designed for batch processing,not real-time detection.To solve this problem,we propose an online fraud detection model using a Neural Factorization Autoencoder(NFA),which analyzes customer calling patterns to detect fraudulent calls.The model employs Neural Factorization Machines(NFM)and an Autoencoder(AE)to model calling patterns and a memory module to adapt to changing customer behaviour.We evaluate our approach on a large dataset of real-world call detail records and compare it with several state-of-the-art methods.Our results show that our approach outperforms the baselines,with an AUC of 91.06%,a TPR of 91.89%,an FPR of 14.76%,and an F1-score of 95.45%.These results demonstrate the effectiveness of our approach in detecting fraud in real-time and suggest that it can be a valuable tool for preventing fraud in telecommunications networks.
文摘文章提出一种在片上系统(System on Chip,SoC)实现高吞吐率的有限状态熵编码(finite state entropy,FSE)算法。通过压缩率、速度、资源消耗、功耗4个方面对所提出的编码器和解码器与典型的硬件哈夫曼编码(Huffman coding,HC)进行性能比较,结果表明,所提出的硬件FSE编码器和解码器具有显著优势。硬件FSE(hFSE)架构实现在SoC的处理系统和可编程逻辑块(programmable logic,PL)上,通过高级可扩展接口(Advanced eXtensible Interface 4,AXI4)总线连接SoC的处理系统和可编程逻辑块。算法测试显示,FSE算法在非均匀数据分布和大数据量情况下,具有更好的压缩率。该文设计的编码器和解码器已在可编程逻辑块上实现,其中包括1个可配置的缓冲模块,将比特流作为单字节或双字节配置输出到8 bit位宽4096深度或16 bit位宽2048深度的块随机访问存储器(block random access memory,BRAM)中。所提出的FSE硬件架构为实时压缩应用提供了高吞吐率、低功耗和低资源消耗的硬件实现。
基金This work was supported by the National Research Foundation of Korea(NRF)Grant(Nos.2018R1A5A7059549,2020R1A2C1014037)by Institute of Information&Communications Technology Planning&Evaluation(IITP)Grant(No.2020-0-01373)funded by the Korea government(*MSIT).*Ministry of Science and Information&Communication Technology.
文摘Much like humans focus solely on object movement to understand actions,directing a deep learning model’s attention to the core contexts within videos is crucial for improving video comprehension.In the recent study,Video Masked Auto-Encoder(VideoMAE)employs a pre-training approach with a high ratio of tube masking and reconstruction,effectively mitigating spatial bias due to temporal redundancy in full video frames.This steers the model’s focus toward detailed temporal contexts.However,as the VideoMAE still relies on full video frames during the action recognition stage,it may exhibit a progressive shift in attention towards spatial contexts,deteriorating its ability to capture the main spatio-temporal contexts.To address this issue,we propose an attention-directing module named Transformer Encoder Attention Module(TEAM).This proposed module effectively directs the model’s attention to the core characteristics within each video,inherently mitigating spatial bias.The TEAM first figures out the core features among the overall extracted features from each video.After that,it discerns the specific parts of the video where those features are located,encouraging the model to focus more on these informative parts.Consequently,during the action recognition stage,the proposed TEAM effectively shifts the VideoMAE’s attention from spatial contexts towards the core spatio-temporal contexts.This attention-shift manner alleviates the spatial bias in the model and simultaneously enhances its ability to capture precise video contexts.We conduct extensive experiments to explore the optimal configuration that enables the TEAM to fulfill its intended design purpose and facilitates its seamless integration with the VideoMAE framework.The integrated model,i.e.,VideoMAE+TEAM,outperforms the existing VideoMAE by a significant margin on Something-Something-V2(71.3%vs.70.3%).Moreover,the qualitative comparisons demonstrate that the TEAM encourages the model to disregard insignificant features and focus more on the essential video features,capturing more detailed spatio-temporal contexts within the video.
