Algorithms Seminar/Fall10
From ResearchWiki
Modelling Data With Uncertainty
Wed 1:25-2:45pm
MEB 3147 (LCR)
Contents |
Synopsis
We will cover many recent developments in the modeling and processing of uncertain data in computer science. The seminar will focus and the modeling, algorithmic, and data-structural foundations needed for understanding, processing, and visualizing uncertain data. First we will overview different models of uncertainty, tracing their origins and motivation (Many-world models in databases, imprecision models in geometry, probabilistic models in databases). Then we will proceed to survey recent developments in computational geometry, databases, visualization, and machine learning and statistics.
- The computational geometry section will describe basic algorithmic tools and analysis several models of uncertain data.
- The databases section will focus on data structures to manage and process large amounts of uncertain data concisely.
- The visualization section will look at models for data uncertainty for surfaces and volumes, and how the resulting data is visualized.
- The machine learning and statistics section will cover classic techniques for learning and detecting uncertainty.
Through this seminar, participants will gain an understanding of the state-of-the-art in modeling and processing uncertain data and will be exposed to several important open problems and exciting research directions.
Participants
- Jeff Phillips, CI Postdoctoral Fellow, School of Computing
- Suresh Venkatasubramanian, Assistant Professor, School of Computing
- Parasaran Raman, PhD Student, School of Computing
- John Moeller, PhD Student, School of Computing
- Avishek Saha, PhD Student, School of Computing
- Seth Juarez, PhD Student, School of Computing
- Kristi Potter, Post-Doc, SCI
- Piyush Rai, PhD Student, School of Computing
Schedule
(subject to change)
| Date | Topic | Outline and Paper(s) | Presenter |
|---|---|---|---|
| Aug 25 | Models of Data Uncertainty | Motivation behind modeling Uncertainty. Survey different models and where they arise. | Jeff Phillips |
| Geometry | |||
| Sep 1 | Range Spaces and epsilon-Samples | How well does random sampling work? Define important geometric-combinatorial notions: Range spaces, eps-nets, eps-samples. Prove random sampling bound for eps-samples (Sariel's Notes read 5.1 and 5.3.1 and 5.3.2 and the "naive proof"). Better deterministic and from continuous to discrete (also a different way of defining concepts). If time, show cool application of eps-nets on uncertain data in sensor nets (eps-Sentinels) | John Moeller |
| Sep 8 | Epsilon-Quantizations and Epsilon-SIPs | Modeling questions on uncertain data with probability distributions. Review specific applications to smallest enclosing ball. (Shape Fitting on Point Sets with Probability Distributions) | Seth Juarez |
| Sep 15 | Geometry on Imprecise Points | Imprecision in data. eps-geometry for rounding errors (eps-Geometry). More general geometry problems: Basic Geometry Measures on Imprecise Points and maybe Largest and Smallest Convex Hulls on Imprecise Points | Parasaran |
| Databases | |||
| Sep 22 | Sketching and Streaming | How to maintain a concise summary of uncertain data on-the-fly. Sketching probabilistic data streams | John Moeller |
| Sep 29 | Histograms | Building Histograms and other representations of uncertain probability distributions. Histograms and wavelets on probabilistic data Probabilistic Histograms for Probabilistic Data | Avishek |
| Oct 6 | Ranking | Retrieving the top (most important) data points, when all values are uncertain. A Unified Approach to Ranking in Probabilistic Databases with slides and Semantics of ranking queries for probabilistic data and expected ranks | Shantharaju |
| Oct 27 | Vis Topic 1 | ||
| Oct 27 | Clustering | Constructing clusters of uncertain data sets. Approximation Algorithms for Clustering Uncertain Data | Parasaran Raman |
| Visualization (see page on Uncertainty Visualization) | |||
| Nov 3 | Vis Topic 2 | ||
| Nov 10 | Uncertain Vector Fields | Visualizing Uncertainty in Vector Fields (2D Vector Field Uncertainty). I will be discussing my project as well. | Harsh Bhatia |
| Machine Learning And Statistics | |||
| Nov 17 | Support Vector Machines | Classifying uncertain data. Support Vector Classification with Input Data Uncertainty | Piyush Rai |
| Nov 24 | Markov Random Fields | Harnessing Locality. | Bo |
| Dec 1 | Particle Filters | Maintaining uncertain estimates. | Harsh |
| Dec 8 | Multi-Armed Bandits | Trading off investigation of uncertainty and rewards. | Avishek |
