We report the world's first all-season training and validation sample sets for global land cover classification with Landsat-8 data.Prior to this,such samples were only available at a single date primarily from th...We report the world's first all-season training and validation sample sets for global land cover classification with Landsat-8 data.Prior to this,such samples were only available at a single date primarily from the growing season.It is unknown how much limitation such a single-date sample has to mapping global land cover in other seasons of the year.To answer this question,we selected available Landsat-8 images from four seasons and collected training and validation samples from them.We compared the performances of training samples in different seasons using Random Forest algorithm.We found that the use of training samples from any individual season would result in the best overall classification accuracy when validated by samples in the same season.The global overall accuracy from combined best seasonal results was 67.2% when classifying the 11 Level-1 classes in the Finer Resolution Observation and Monitoring of Global Land Cover(FROM-GLC) classification system.The use of training samples from all seasons(named all-season training sample set hereafter) produced an overall accuracy of 67.0%.We also tested classification within 10° latitude 60° longitude zones using all-season training subsample within each zone and obtained an overall accuracy of 70.2%.This indicates that properly grouped subsamples in space can help improve classification accuracies.All the results in this study seem to suggest that it is possible to use an all-season training sample set to reach global optimality with universal applicability in classifying images acquired at any time of a year for global land cover mapping.展开更多
A variable dimensional state space（VDSS） has been proposed to improve the re-planning time when the robotic systems operate in large unknown environments.VDSS is constructed by uniforming lattice state space and gri...A variable dimensional state space（VDSS） has been proposed to improve the re-planning time when the robotic systems operate in large unknown environments.VDSS is constructed by uniforming lattice state space and grid state space.In VDSS,the lattice state space is only used to construct search space in the local area which is a small circle area near the robot,and grid state space elsewhere.We have tested VDSS with up to 80 indoor and outdoor maps in simulation and on segbot robot platform.Through the simulation and segbot robot experiments,it shows that exploring on VDSS is significantly faster than exploring on lattice state space by Anytime Dynamic A＊（AD＊） planner and VDSS is feasible to be used on robotic systems.展开更多
In many planning situations, computation itself becomes a resource to be planned and scheduled. We model such computational resources as conventional resources which are used by control-flow actions, e.g., to direc...In many planning situations, computation itself becomes a resource to be planned and scheduled. We model such computational resources as conventional resources which are used by control-flow actions, e.g., to direct the planning process. Control-flow actions and conventional actions are planned/scheduled in an integrated way and can interact with each other. Control-flow actions are then executed by the planning engine itself. The approach is illustrated by examples, e.g., for hierarchical planning, in which tasks that are temporally still far away impose only rough constraints on the current schedule, and control-flow tasks ensure that these tasks are refined as they approach the current time. Using the same mechanism, anytime algorithms can change appropriate search methods or parameters over time, and problems like scheduling critical time-outs for garbage collection can be made part of the planning itself.展开更多
Classification,using the decision tree algorithm,is a widely studied problem in data streams.The challenge is when to split a decision node into multiple leaves.Concentration inequalities,that exploit variance informa...Classification,using the decision tree algorithm,is a widely studied problem in data streams.The challenge is when to split a decision node into multiple leaves.Concentration inequalities,that exploit variance information such as Bernstein's and Bennett's inequalities,are often substantially strict as compared with Hoeffding's bound which disregards variance.Many machine learning algorithms for stream classification such as very fast decision tree(VFDT) learner,AdaBoost and support vector machines(SVMs),use the Hoeffding's bound as a performance guarantee.In this paper,we propose a new algorithm based on the recently proposed empirical Bernstein's bound to achieve a better probabilistic bound on the accuracy of the decision tree.Experimental results on four synthetic and two real world data sets demonstrate the performance gain of our proposed technique.展开更多
基金partially supported by the National High Technology Program(2013AA122804)the Special Fund for Meteorology Scientific Research in the Public Welfare(GYHY201506023)of ChinaOpen Fund of State Key Laboratory of Remote Sensing Science(OFSLRSS201514)
文摘We report the world's first all-season training and validation sample sets for global land cover classification with Landsat-8 data.Prior to this,such samples were only available at a single date primarily from the growing season.It is unknown how much limitation such a single-date sample has to mapping global land cover in other seasons of the year.To answer this question,we selected available Landsat-8 images from four seasons and collected training and validation samples from them.We compared the performances of training samples in different seasons using Random Forest algorithm.We found that the use of training samples from any individual season would result in the best overall classification accuracy when validated by samples in the same season.The global overall accuracy from combined best seasonal results was 67.2% when classifying the 11 Level-1 classes in the Finer Resolution Observation and Monitoring of Global Land Cover(FROM-GLC) classification system.The use of training samples from all seasons(named all-season training sample set hereafter) produced an overall accuracy of 67.0%.We also tested classification within 10° latitude 60° longitude zones using all-season training subsample within each zone and obtained an overall accuracy of 70.2%.This indicates that properly grouped subsamples in space can help improve classification accuracies.All the results in this study seem to suggest that it is possible to use an all-season training sample set to reach global optimality with universal applicability in classifying images acquired at any time of a year for global land cover mapping.
基金Supported by the National Natural Science Foundation of China(90920304)
文摘A variable dimensional state space（VDSS） has been proposed to improve the re-planning time when the robotic systems operate in large unknown environments.VDSS is constructed by uniforming lattice state space and grid state space.In VDSS,the lattice state space is only used to construct search space in the local area which is a small circle area near the robot,and grid state space elsewhere.We have tested VDSS with up to 80 indoor and outdoor maps in simulation and on segbot robot platform.Through the simulation and segbot robot experiments,it shows that exploring on VDSS is significantly faster than exploring on lattice state space by Anytime Dynamic A＊（AD＊） planner and VDSS is feasible to be used on robotic systems.
文摘In many planning situations, computation itself becomes a resource to be planned and scheduled. We model such computational resources as conventional resources which are used by control-flow actions, e.g., to direct the planning process. Control-flow actions and conventional actions are planned/scheduled in an integrated way and can interact with each other. Control-flow actions are then executed by the planning engine itself. The approach is illustrated by examples, e.g., for hierarchical planning, in which tasks that are temporally still far away impose only rough constraints on the current schedule, and control-flow tasks ensure that these tasks are refined as they approach the current time. Using the same mechanism, anytime algorithms can change appropriate search methods or parameters over time, and problems like scheduling critical time-outs for garbage collection can be made part of the planning itself.
基金the National Natural Science Foundation of China(Nos.60873108,61175047 and 61152001)the Fundamental Research Funds for the Central Universities of China(No.SWJTU11ZT08)
文摘Classification,using the decision tree algorithm,is a widely studied problem in data streams.The challenge is when to split a decision node into multiple leaves.Concentration inequalities,that exploit variance information such as Bernstein's and Bennett's inequalities,are often substantially strict as compared with Hoeffding's bound which disregards variance.Many machine learning algorithms for stream classification such as very fast decision tree(VFDT) learner,AdaBoost and support vector machines(SVMs),use the Hoeffding's bound as a performance guarantee.In this paper,we propose a new algorithm based on the recently proposed empirical Bernstein's bound to achieve a better probabilistic bound on the accuracy of the decision tree.Experimental results on four synthetic and two real world data sets demonstrate the performance gain of our proposed technique.
