Since the fully convolutional network has achieved great success in semantic segmentation,lots of works have been proposed to extract discriminative pixel representations.However,the authors observe that existing meth...Since the fully convolutional network has achieved great success in semantic segmentation,lots of works have been proposed to extract discriminative pixel representations.However,the authors observe that existing methods still suffer from two typical challenges:(i)The intra-class feature variation between different scenes may be large,leading to the difficulty in maintaining the consistency between same-class pixels from different scenes;(ii)The inter-class feature distinction in the same scene could be small,resulting in the limited performance to distinguish different classes in each scene.The authors first rethink se-mantic segmentation from a perspective of similarity between pixels and class centers.Each weight vector of the segmentation head represents its corresponding semantic class in the whole dataset,which can be regarded as the embedding of the class center.Thus,the pixel-wise classification amounts to computing similarity in the final feature space between pixels and the class centers.Under this novel view,the authors propose a Class Center Similarity(CCS)layer to address the above-mentioned challenges by generating adaptive class centers conditioned on each scenes and supervising the similarities between class centers.The CCS layer utilises the Adaptive Class Center Module to generate class centers conditioned on each scene,which adapt the large intra-class variation between different scenes.Specially designed Class Distance Loss(CD Loss)is introduced to control both inter-class and intra-class distances based on the predicted center-to-center and pixel-to-center similarity.Finally,the CCS layer outputs the processed pixel-to-center similarity as the segmentation prediction.Extensive experiments demonstrate that our model performs favourably against the state-of-the-art methods.展开更多
The on-demand food delivery(OFD)service has gained rapid development in the past decades but meanwhile encounters challenges for further improving operation quality.The order dispatching problem is one of the most con...The on-demand food delivery(OFD)service has gained rapid development in the past decades but meanwhile encounters challenges for further improving operation quality.The order dispatching problem is one of the most concerning issues for the OFD platforms,which refer to dynamically dispatching a large number of orders to riders reasonably in very limited decision time.To solve such a challenging combinatorial optimization problem,an effective matching algorithm is proposed by fusing the reinforcement learning technique and the optimization method.First,to deal with the large-scale complexity,a decoupling method is designed by reducing the matching space between new orders and riders.Second,to overcome the high dynamism and satisfy the stringent requirements on decision time,a reinforcement learning based dispatching heuristic is presented.To be specific,a sequence-to-sequence neural network is constructed based on the problem characteristic to generate an order priority sequence.Besides,a training approach is specially designed to improve learning performance.Furthermore,a greedy heuristic is employed to effectively dispatch new orders according to the order priority sequence.On real-world datasets,numerical experiments are conducted to validate the effectiveness of the proposed algorithm.Statistical results show that the proposed algorithm can effectively solve the problem by improving delivery efficiency and maintaining customer satisfaction.展开更多
On-demand food delivery(OFD)is gaining more and more popularity in modern society.As a kernel order assignment manner in OFD scenario,order recommendation directly influences the delivery efficiency of the platform an...On-demand food delivery(OFD)is gaining more and more popularity in modern society.As a kernel order assignment manner in OFD scenario,order recommendation directly influences the delivery efficiency of the platform and the delivery experience of riders.This paper addresses the dynamism of the order recommendation problem and proposes a reinforcement learning solution method.An actor-critic network based on long short term memory(LSTM)unit is designed to deal with the order-grabbing conflict between different riders.Besides,three rider sequencing rules are accordingly proposed to match different time steps of the LSTM unit with different riders.To test the performance of the proposed method,extensive experiments are conducted based on real data from Meituan delivery platform.The results demonstrate that the proposed reinforcement learning based order recommendation method can significantly increase the number of grabbed orders and reduce the number of order-grabbing conflicts,resulting in better delivery efficiency and experience for the platform and riders.展开更多
Low-cost Global Navigation Satellite System(GNSS)devices offer a cost-effective alternative to traditional GNSS systems,making GNSS technology accessible to a wider range of applications.Nevertheless,low-cost GNSS dev...Low-cost Global Navigation Satellite System(GNSS)devices offer a cost-effective alternative to traditional GNSS systems,making GNSS technology accessible to a wider range of applications.Nevertheless,low-cost GNSS devices often face the challenges in effectively capturing and tracking satellite signals,which leads to losing the observations at certain frequencies.Moreover,the observation peculiarities of low-cost devices are in contradistinction to those of traditional geodetic GNSS receivers.In this contribution,a low-cost PPP-RTK model that considers the unique characteristics of different types of measurements is developed and its performance is fully evaluated with u-blox F9P receivers equipped with three distinctive antenna configurations:vertical dipole,microstrip patch,and helix antennas.Several static and kinematic experiments in different scenarios are conducted to verify the effectiveness of the proposed method.The results indicate that the mixed-frequency PPP-RTK model outperforms the traditional dual-frequency one with higher positioning accuracy and fixing percentage.Among the three low-cost antennas tested,the vertical dipole antenna demonstrates the best performance under static conditions and shows a comparable performance as geodetic antennas with a positioning accuracy of 0.02 m,0.01 m and 0.07 m in the east,north,and up components,respectively.Under low-speed kinematic scenarios,the helix antenna outperforms the other two with a positioning accuracy of(0.07 m,0.07 m,0.34 m).Furthermore,the helix antenna is also proved to be the best choice for vehicle navigation with an ambiguity fixing rate of over 95%and a positioning accuracy of(0.13 m,0.14 m,0.36 m).展开更多
Knowlege is important for text-related applications.In this paper,we introduce Microsoft Concept Graph,a knowledge graph engine that provides concept tagging APIs to facilitate the understanding of human languages.Mic...Knowlege is important for text-related applications.In this paper,we introduce Microsoft Concept Graph,a knowledge graph engine that provides concept tagging APIs to facilitate the understanding of human languages.Microsoft Concept Graph is built upon Probase,a universal probabilistic taxonomy consisting of instances and concepts mined from the Web.We start by introducing the construction of the knowledge graph through iterative semantic extraction and taxonomy construction procedures,which extract 2.7 million concepts from 1.68 billion Web pages.We then use conceptualization models to represent text in the concept space to empower text-related applications,such as topic search,query recommendation,Web table understanding and Ads relevance.Since the release in 2016,Microsoft Concept Graph has received more than 100,000 pageviews,2 million API calls and 3,000 registered downloads from 50,000 visitors over 64 countries.展开更多
Motivated by the practice that e-sellers cooperate with insurance companies to offer consumers the return-freight insurance(RI),this paper aims to investigate the competing e-sellers’RI strategies.Regarding the infor...Motivated by the practice that e-sellers cooperate with insurance companies to offer consumers the return-freight insurance(RI),this paper aims to investigate the competing e-sellers’RI strategies.Regarding the information asymmetry in the online context,reputation system is widely applied by e-platforms.In an online market with two competing e-sellers that sell the same product but are differentiated in their reputation,this paper builds an analytical model to explore the e-sellers optimal pricing and RI strategies.Combined with sellers’reputation and their RI strategies,the equilibrium outcomes under four cases are discussed.This paper reveals the conditions that e-sellers are willing to offer RI.Specifically,the findings demonstrate that low reputation e-seller is more likely to offer RI.Moreover,when the sellers are more divergent,they are more likely to co-exist in the market.Insurance premium and RI compensation play critical roles in their decisions.RI introduction tends to increase the price,thus offsets the benefits of RI,but does not affect the total consumer surplus.展开更多
基金Hubei Provincial Natural Science Foundation of China,Grant/Award Number:2022CFA055National Natural Science Foundation of China,Grant/Award Number:62176097。
文摘Since the fully convolutional network has achieved great success in semantic segmentation,lots of works have been proposed to extract discriminative pixel representations.However,the authors observe that existing methods still suffer from two typical challenges:(i)The intra-class feature variation between different scenes may be large,leading to the difficulty in maintaining the consistency between same-class pixels from different scenes;(ii)The inter-class feature distinction in the same scene could be small,resulting in the limited performance to distinguish different classes in each scene.The authors first rethink se-mantic segmentation from a perspective of similarity between pixels and class centers.Each weight vector of the segmentation head represents its corresponding semantic class in the whole dataset,which can be regarded as the embedding of the class center.Thus,the pixel-wise classification amounts to computing similarity in the final feature space between pixels and the class centers.Under this novel view,the authors propose a Class Center Similarity(CCS)layer to address the above-mentioned challenges by generating adaptive class centers conditioned on each scenes and supervising the similarities between class centers.The CCS layer utilises the Adaptive Class Center Module to generate class centers conditioned on each scene,which adapt the large intra-class variation between different scenes.Specially designed Class Distance Loss(CD Loss)is introduced to control both inter-class and intra-class distances based on the predicted center-to-center and pixel-to-center similarity.Finally,the CCS layer outputs the processed pixel-to-center similarity as the segmentation prediction.Extensive experiments demonstrate that our model performs favourably against the state-of-the-art methods.
基金supported in part by the National Natural Science Foundation of China(No.62273193)Tsinghua University-Meituan Joint Institute for Digital Life,and the Research and Development Project of CRSC Research&Design Institute Group Co.,Ltd.
文摘The on-demand food delivery(OFD)service has gained rapid development in the past decades but meanwhile encounters challenges for further improving operation quality.The order dispatching problem is one of the most concerning issues for the OFD platforms,which refer to dynamically dispatching a large number of orders to riders reasonably in very limited decision time.To solve such a challenging combinatorial optimization problem,an effective matching algorithm is proposed by fusing the reinforcement learning technique and the optimization method.First,to deal with the large-scale complexity,a decoupling method is designed by reducing the matching space between new orders and riders.Second,to overcome the high dynamism and satisfy the stringent requirements on decision time,a reinforcement learning based dispatching heuristic is presented.To be specific,a sequence-to-sequence neural network is constructed based on the problem characteristic to generate an order priority sequence.Besides,a training approach is specially designed to improve learning performance.Furthermore,a greedy heuristic is employed to effectively dispatch new orders according to the order priority sequence.On real-world datasets,numerical experiments are conducted to validate the effectiveness of the proposed algorithm.Statistical results show that the proposed algorithm can effectively solve the problem by improving delivery efficiency and maintaining customer satisfaction.
基金supported in part by the National Natural Science Foundation of China(No.62273193)Tsinghua University-Meituan Joint Institute for Digital Life,and the Research and Development Project of CRSC Research&Design Institute Group Co.,Ltd.
文摘On-demand food delivery(OFD)is gaining more and more popularity in modern society.As a kernel order assignment manner in OFD scenario,order recommendation directly influences the delivery efficiency of the platform and the delivery experience of riders.This paper addresses the dynamism of the order recommendation problem and proposes a reinforcement learning solution method.An actor-critic network based on long short term memory(LSTM)unit is designed to deal with the order-grabbing conflict between different riders.Besides,three rider sequencing rules are accordingly proposed to match different time steps of the LSTM unit with different riders.To test the performance of the proposed method,extensive experiments are conducted based on real data from Meituan delivery platform.The results demonstrate that the proposed reinforcement learning based order recommendation method can significantly increase the number of grabbed orders and reduce the number of order-grabbing conflicts,resulting in better delivery efficiency and experience for the platform and riders.
基金National Natural Science Foundation of China,41974027,Xingxing Li42204017,Xin Li+2 种基金National Postdoctoral Program for Innovative Talents,China,BX20220239,Xin Lithe special fund of Hubei Luojia Laboratory,220100006,Xin Lithe Fundamental Research Funds for the Central Universities,2042022kf1001,Xin Li.
文摘Low-cost Global Navigation Satellite System(GNSS)devices offer a cost-effective alternative to traditional GNSS systems,making GNSS technology accessible to a wider range of applications.Nevertheless,low-cost GNSS devices often face the challenges in effectively capturing and tracking satellite signals,which leads to losing the observations at certain frequencies.Moreover,the observation peculiarities of low-cost devices are in contradistinction to those of traditional geodetic GNSS receivers.In this contribution,a low-cost PPP-RTK model that considers the unique characteristics of different types of measurements is developed and its performance is fully evaluated with u-blox F9P receivers equipped with three distinctive antenna configurations:vertical dipole,microstrip patch,and helix antennas.Several static and kinematic experiments in different scenarios are conducted to verify the effectiveness of the proposed method.The results indicate that the mixed-frequency PPP-RTK model outperforms the traditional dual-frequency one with higher positioning accuracy and fixing percentage.Among the three low-cost antennas tested,the vertical dipole antenna demonstrates the best performance under static conditions and shows a comparable performance as geodetic antennas with a positioning accuracy of 0.02 m,0.01 m and 0.07 m in the east,north,and up components,respectively.Under low-speed kinematic scenarios,the helix antenna outperforms the other two with a positioning accuracy of(0.07 m,0.07 m,0.34 m).Furthermore,the helix antenna is also proved to be the best choice for vehicle navigation with an ambiguity fixing rate of over 95%and a positioning accuracy of(0.13 m,0.14 m,0.36 m).
文摘Knowlege is important for text-related applications.In this paper,we introduce Microsoft Concept Graph,a knowledge graph engine that provides concept tagging APIs to facilitate the understanding of human languages.Microsoft Concept Graph is built upon Probase,a universal probabilistic taxonomy consisting of instances and concepts mined from the Web.We start by introducing the construction of the knowledge graph through iterative semantic extraction and taxonomy construction procedures,which extract 2.7 million concepts from 1.68 billion Web pages.We then use conceptualization models to represent text in the concept space to empower text-related applications,such as topic search,query recommendation,Web table understanding and Ads relevance.Since the release in 2016,Microsoft Concept Graph has received more than 100,000 pageviews,2 million API calls and 3,000 registered downloads from 50,000 visitors over 64 countries.
基金the National Natural Science Foundation of China(71971165)the National Key Research and Development Program of China(2021YFB3301801)+1 种基金the MOE Project of Humanities and Social Science of China(19YJE630002)the Soft Science Research Program of Shannxi(2018KRZ005)。
文摘Motivated by the practice that e-sellers cooperate with insurance companies to offer consumers the return-freight insurance(RI),this paper aims to investigate the competing e-sellers’RI strategies.Regarding the information asymmetry in the online context,reputation system is widely applied by e-platforms.In an online market with two competing e-sellers that sell the same product but are differentiated in their reputation,this paper builds an analytical model to explore the e-sellers optimal pricing and RI strategies.Combined with sellers’reputation and their RI strategies,the equilibrium outcomes under four cases are discussed.This paper reveals the conditions that e-sellers are willing to offer RI.Specifically,the findings demonstrate that low reputation e-seller is more likely to offer RI.Moreover,when the sellers are more divergent,they are more likely to co-exist in the market.Insurance premium and RI compensation play critical roles in their decisions.RI introduction tends to increase the price,thus offsets the benefits of RI,but does not affect the total consumer surplus.