Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Associating earth-orbiting objects detected by astronomical telescopes Haseena Ahmed Prince Chidyagwai Kun Gou Yun Liu Timur Milgrom Vincent Quenneville-Bélair Mentor: Dr. Gary B. Green (The Aerospace Corporation) August 17, 2007 Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Introduction Problem Statement Streak Modeling Solution Approach - Cluster Analysis Agglomerative Hierarchical Clustering k-means Clustering Comparison Implementation and Results Conclusions Future Work Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Problem Statement Streak Modeling Problem Statement Satellites make streaks in telescope images I Input: 1. Streak data 2. Orbit data I Objective: Identify streaks made by the same object I Process: Take the image and find the streak (astronomers) Estimate the orbit of the object (orbit analysts) Cluster streaks (large cardinality problem - our task) Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Problem Statement Streak Modeling Streak Modeling Streaks can be modeled in two spaces: I Image space: A vector in R3 as a result of processing streak points points in a streak {RAi , DEi , ti }#of → {RA, DE , α} i=1 I Orbit space: A vector in R6 as a result of orbit estimation points in a streak → {a, e, i, Ω, ωp , M} {RAi , DEi , ti }#of i=1 Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Clustering • Similarity and dissimilarity measures • Depends mainly on the data set available Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Clustering • Similarity and dissimilarity measures • Depends mainly on the data set available • Two commonly used methods of clustering I Hierarchical clustering I I Tree structure Agglomerative Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Clustering • Similarity and dissimilarity measures • Depends mainly on the data set available • Two commonly used methods of clustering I Hierarchical clustering I I I Tree structure Agglomerative Partitional clustering I I One level partitioning k-means Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Agglomerative Hierarchical Clustering Given a set of points to be clustered in 2D as in the figure I We need to specify: distance measure, type of linkage Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Agglomerative Hierarchical Clustering Compute the proximity matrix as in table Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Agglomerative Hierarchical Clustering I Cluster points 3 and 6 I Obtain new proximity matrix by calculating the distance between the new cluster {3, 6} and other points dist({3, 6} , {1}) = min (dist(3, 1), dist(6, 1)) = min (0.22, 0.23) = 0.22 Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Agglomerative Hierarchical Clustering Dendogram representation can be given by figure Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison k-means Clustering Algorithm: I Select k points as initial centroids I repeat Form k clusters by assigning each point to closest centroid. Recompute the centroid of each cluster. until Centroids do not change. Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Comparision - Agglomerative vs k-means I Agglomerative I I I Complexity is O(n2 ) in memory and O(n2 log n) in CPU time Local optimal clustering All merges are final Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Agglomerative Hierarchical Clustering k-means Clustering Comparison Comparision - Agglomerative vs k-means I Agglomerative I I I I Complexity is O(n2 ) in memory and O(n2 log n) in CPU time Local optimal clustering All merges are final k-means I I I I Complexity is O(n) in memory space and CPU time Number of clusters k needs to be known a-priori Initialization of centers of clusters Local optimal clustering Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Implementation I Representations in orbit space I I I I Kepler (Orbit space) Equinoctial elements Cartesian ellipse MATLAB I I Linkage Distance function Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Time comparision Satellites 6 32 74 137 Streaks 96 861 2191 4086 Kepler Time .05 3.85 56.45 423.13 Ellipse Time .06 4.48 61.7 443.17 Table: Computational time (seconds) Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Silhouette For unperturbed data Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Silhouette For perturbed data Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Distance Measure Comparison Satellites 6 36 74 137 Euclidean 63 612 1563 3107 Weighted 7 99 273 764 Cosine 644 617 1537 3098 Table: Performance of norms (# clusters) Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Linkage Function Comparison Satellites 6 36 74 137 Single 7 99 273 764 Average 13 86 260 520 Centroid 13 82 240 472 Table: Performance of linkage (# clusters) Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Effect of Variation in Cut-off Satellites 6 36 36 74 137 Found 6 32 33 57 133 Cut-off 1.154 1.1546 1.1547 1.1546331 1.1546 Silhouette 0.70 0.70 0.79 0.48 0.47 Table: Effect of cut-off on silhouette (a, e weighted with 0.1) Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Large data clustering Sectioning method tested on I Number of streaks = 4400 I Actual number of satellites = 137 Sections Time Found 1 356 137 2 143 116 4 56 126 8 12 143 Table: Effective grouping Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Conclusions I Weighted norm is effective I Linkage function is inconclusive I Cut-off is sensitive I Sectional method is promising Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Future Work Improving clustering I Develop theory for choosing weights I Develop theory for choosing cutoff Improving sectioning method I Optimal grouping Team 3 Associating Earth-Orbiting Objects Outline Introduction Solution Approach - Cluster Analysis Implementation and Results Conclusions Future Work Questions? Team 3 Associating Earth-Orbiting Objects