Visualization and Cluster

Slides:



Advertisements
Similar presentations
Pseudo-Relevance Feedback For Multimedia Retrieval By Rong Yan, Alexander G. and Rong Jin Mwangi S. Kariuki
Advertisements

Improvements and extras Paul Thomas CSIRO. Overview of the lectures 1.Introduction to information retrieval (IR) 2.Ranked retrieval 3.Probabilistic retrieval.
Node-Attribute Graph Layout for Small-World Networks Helen Gibson Principal Supervisor: Dr. Paul Vickers 1 st Supervisor: Dr. Maia Angelova 2 nd Supervisor:
Aggregating local image descriptors into compact codes
Computer vision: models, learning and inference Chapter 8 Regression.
PolyAnalyst Data and Text Mining tool Your Knowledge Partner TM www
VALTChessVA IntroAppsWrap-up 1/25 User-Centric Visual Analytics Remco Chang Tufts University Department of Computer Science.
Universal Search and Social Networking Exploiting the features of each to enhance the other and the tools that make it possible Peter Wallqvist Ravn Systems.
Visual Analytics Research at WPI Dr. Matthew Ward and Dr. Elke Rundensteiner Computer Science Department.
Computer Vision Spring ,-685 Instructor: S. Narasimhan Wean 5403 T-R 3:00pm – 4:20pm Lecture #20.
One-Shot Multi-Set Non-rigid Feature-Spatial Matching
ETHEM ALPAYDIN © The MIT Press, Lecture Slides for.
CS 589 Information Risk Management 6 February 2007.
Dimensionality Reduction with Linear Transformations project update by Mingyue Tan March 17, 2004.
Dimensionality Reduction
WPI Center for Research in Exploratory Data and Information Analysis CREDIA SC4DEVO-1, July 12-15, 2004 Interactive Visual Exploration of Multivariate.
Chapter 5: Query Operations Baeza-Yates, 1999 Modern Information Retrieval.
Multivariate Data Visualization Adapted from Slides by: Matthew O. Ward Computer Science Department Worcester Polytechnic Institute This work was supported.
Relevance Feedback based on Parameter Estimation of Target Distribution K. C. Sia and Irwin King Department of Computer Science & Engineering The Chinese.
Prénom Nom Document Analysis: Data Analysis and Clustering Prof. Rolf Ingold, University of Fribourg Master course, spring semester 2008.
The Table Lens: Merging Graphical and Symbolic Representations in an Interactive Focus + Context Visualization for Tabular Information R. Rao and S. K.
Intelligent User Interfaces Research Group Directed by: Frank Shipman.
LabVis: Simple and Intuitive Visualization of Medical Laboratory Results Adam Bodnar and Dmitry Nekrasovski CPSC 533 Final Project Presentation April 21,
Paper Summary of: Modelling Retrieval and Navigation in Context by: Massimo Melucci Ahmed A. AlNazer May 2008 ICS-542: Multimedia Computing – 072.
Visual Recognition Tutorial
INTRODUCTION TO Machine Learning ETHEM ALPAYDIN © The MIT Press, Lecture Slides for.
The Visual Knowledge Builder: A Second Generation Spatial Hypertext Frank M. Shipman III Haowei Hsieh Preetam Maloor J. Michael Moore.
Memoplex Browser: Searching and Browsing in Semantic Networks CPSC 533C - Project Update Yoel Lanir.
Knowledge Science & Engineering Institute, Beijing Normal University, Analyzing Transcripts of Online Asynchronous.
© Ramesh Jain Ramesh Jain CTO, PRAJA inc. and Professor Emeritus, UCSD Emergent Semantics and Experiential Computing.
The Tutorial of Principal Component Analysis, Hierarchical Clustering, and Multidimensional Scaling Wenshan Wang.
Chapter 3 Data Exploration and Dimension Reduction 1.
Dist FuncIntroPersonalityProvenanceGroupWrap-up 1/40 User-Centric Visual Analytics Remco Chang Tufts University.
1 A Bayesian Method for Guessing the Extreme Values in a Data Set Mingxi Wu, Chris Jermaine University of Florida September 2007.
What are your interactions doing for your visualization? Remco Chang UNC Charlotte Charlotte Visualization Center.
Taxonomies of Visualization Techniques CMPT 455/826 - Week 12, Day 2 w12d2 Sept-Dec
Visual Perspectives iPLANT Visual Analytics Workshop November 5-6, 2009 ;lk Visual Analytics Bernice Rogowitz Greg Abram.
Enhancing Interactive Visual Data Analysis by Statistical Functionality Jürgen Platzer VRVis Research Center Vienna, Austria.
Implicit User Feedback Hongning Wang Explicit relevance feedback 2 Updated query Feedback Judgments: d 1 + d 2 - d 3 + … d k -... Query User judgment.
Learning Geographical Preferences for Point-of-Interest Recommendation Author(s): Bin Liu Yanjie Fu, Zijun Yao, Hui Xiong [KDD-2013]
Toward A Session-Based Search Engine Smitha Sriram, Xuehua Shen, ChengXiang Zhai Department of Computer Science University of Illinois, Urbana-Champaign.
1 CSC 594 Topics in AI – Text Mining and Analytics Fall 2015/16 7. Topic Extraction.
Relevance Feedback Hongning Wang What we have learned so far Information Retrieval User results Query Rep Doc Rep (Index) Ranker.
Visual Analytics with Linked Open Data and Social Media for e- Governance Vitaveska Lanfranchi Suvodeep Mazumdar Tomi Kauppinen Anna Lisa Gentile Updated.
Indirect Supervision Protocols for Learning in Natural Language Processing II. Learning by Inventing Binary Labels This work is supported by DARPA funding.
Semantic Wordfication of Document Collections Presenter: Yingyu Wu.
Personalized Interaction With Semantic Information Portals Eric Schwarzkopf DFKI
CS3041 – Final week Today: Searching and Visualization Friday: Software tools –Study guide distributed (in class only) Monday: Social Imps –Study guide.
Data Visualization Michel Bruley Teradata Aster EMEA Marketing Director April 2013 Michel Bruley Teradata Aster EMEA Marketing Director.
CSC2535: Computation in Neural Networks Lecture 12: Non-linear dimensionality reduction Geoffrey Hinton.
Implicit User Feedback Hongning Wang Explicit relevance feedback 2 Updated query Feedback Judgments: d 1 + d 2 - d 3 + … d k -... Query User judgment.
Radial Basis Function ANN, an alternative to back propagation, uses clustering of examples in the training set.
A Novel Visualization Model for Web Search Results Nguyen T, and Zhang J IEEE Transactions on Visualization and Computer Graphics PAWS Meeting Presented.
Relevance Feedback Hongning Wang
Multimedia Analytics Jianping Fan Department of Computer Science University of North Carolina at Charlotte.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
Advanced Gene Selection Algorithms Designed for Microarray Datasets Limitation of current feature selection methods: –Ignores gene/gene interaction: single.
1 Database Systems Group Research Overview OLAP Statistical Tests Goal: Isolate factors that cause significant changes in a measured value – Ex:
Jianping Fan Department of Computer Science University of North Carolina at Charlotte Charlotte, NC Relevance Feedback for Image Retrieval.
Clustering Approaches Ka-Lok Ng Department of Bioinformatics Asia University.
Computational Intelligence: Methods and Applications Lecture 26 Density estimation, Expectation Maximization. Włodzisław Duch Dept. of Informatics, UMK.
1 Dongheng Sun 04/26/2011 Learning with Matrix Factorizations By Nathan Srebro.
Personalized Social Image Recommendation
Group Y Presenters: (indicate roles)
Relevance Feedback Hongning Wang
Image Segmentation Techniques
What Visualization can do for Data Clustering?
Matching Words with Pictures
Ying Dai Faculty of software and information science,
Ying Dai Faculty of software and information science,
Presentation transcript:

Visualization and Cluster

Visual Analysis “the science if analytical reasoning facilitated by interactive visual interface” (Thomas and Cook, 2005) Interact with data Test hypotheses Formulate knowledge Human intuitive is reliable Few user are well-versed in algorithms

Clustering without Human User have domain knowledge for feature selection Which feature is more important than which Sometimes feature have different weight in different use scenarios Priority Distribution User know outliers in dataset User can give initial state

Show Cluster with Visualization

BaVA: Approach Bayesian Visual Analytics framework Display posterior result Dimension Reduction for display Expert give feedback through adjust layout Feedback as observation level rather than dimension level What kind of interaction should be captured Human input change the underlying probabilistic model and updates the display

Semantic Layout ForceSPIRE https://www.youtube.com/watch?v=I3cKKSFnePo&feature=youtu.be

Interface design Move Document link document: Pin: semantic location Update position based on current weight link document: need update the weight Pin: semantic location Exp: move document close to a pin is expressive movement Text highlighting Increase the weight for highlighted keywords and update the layout

Interface design Search Document Coloring Visual level of detail Document contain the keywords will be highlighted Document Coloring Mark document of same group Visual level of detail More detail = easier to reference Annotation Give semantic information to clusters = easier to reference

Interaction Feedback Spatial information on interaction is ambiguity Will cover it later Operations are on observation level, not on dimension level Let user adjust parameters directly is just guessing game Also, operation is in 2D space rather than high dimension space

Interaction Feedback Two examples: PPCA and MDS

Probabilistic PCA PCA: minimize the variance e Problem of PCA: Important structures in data may not correlate with variance, like cluster Probabilistic PCA

Probabilistic PCA Let , the marginal variance of d Sd the empirical variance of d MAP(∑d) = Sd Let The coordinate The relationship between variance and coordinate

User guided PPCA Display show point layout in 2D space Drag away or drag close two observations if user thinks they are close

User guided PPCA Dragged points have different and similar features A hypothetical variance matrix f(p) An addition weight v to show how important this adjust is Parameter feedback

User guided PPCA https://www.youtube.com/watch?v=k5NYZbP4yKQ

Weighted MDS Minimize the difference if sum of distance in the real space and in the embedded space

User guided MDS

User guided MDS User adjust the relative position of points Solve w so that r: adjust position, d: original position

V2PI Visual to parametric Interaction Spatial Interaction is intuitive but ambiguity Move a point can means: Move toward to a unmoved points Move around and happen to be moved closer to unmoved points Need better explicit interaction Use tag to distinguish

Interaction Design For each interaction, data are involved in two ways Explicit Implicit Example: move A to B, A is Explicitly involved and B is implicitly involved For unmoved points: It is implicitly involved Not involved at all (ignored by user) Get meaning od unmoved point with minimal addition effort

Interaction Design Moved set Highlighted set (user highlight some points that he found interesting) Untouched set (ignored by user)

Pair-wise weighting C_ij show user’s preference on each data C_ij: Combine value is 1

Pair-wise weighting n_h and n_m are the number of object of highlighted and moved in the dataset

User guided DR Key notes: We can rely on user’s domain knowledge to improve Dimension Reduction Data manipulating must be intuitive and efficient