research overview piyush kumar. 1.optimization 2.computational geometry 3.computer graphics...

12
Research Overview Piyush Kumar

Upload: grace-ryan

Post on 13-Dec-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Research Overview

Piyush Kumar

Page 2: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

1. Optimization

2. Computational Geometry

3. Computer Graphics

4. Pattern Recognition and Machine Learning / Biometrics

5. Robotics

Research Interests

Page 3: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Covering Problems

1. Cover a given point set:

1. With k-shapes (k < 5)

2. With smallest ball

3. With smallest 1-ellipsoid

2. Cover a set of balls/ellipsoids

1. With smallest ball/ellipsoid

Optimization

Page 4: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Covering Problems : An Example

1. Cover a given point set:

1. With k-shapes (k < 5)

Optimization

Page 5: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Covering Problems: Motivation

Clustering for various applications

1. 3D Terrain Covers/ Surveillance

2. Automatic Surface estimation

3. Antenna/Other placement problems.

4. Operations Research

5. Signal/Image Processing.

6. Bio-Metrics / Bio-Informatics

7. Compression.

Page 6: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Computational Geometry: Surface Estimation

1. LIDAR / Laser Range scanner Data Problems

1. How to do surface estimation

2. How to handle massive data sizes

3. How to make sense of the data

Page 7: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Surface estimation problems

1. Other problems

1. How to handle noise

2. How to use SMP machines for fast processing

3. How to do processing online/dynamically

Page 8: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Triangulations

Compute Delaunay Triangulations in 2D for massive data

sets

1. External Memory

2. Parallel

3. Cache-Oblivious (In Theory)

4. Practical Variant

Joint work with E. Ramos. (UIUC)

Courtesy D. Eppstein

Acknowledgements:

Cobalt: Altix at UIUC (1024p) (NCSA, UIUC) Gestalt: 4P Opteron at CAVIS.

Page 9: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Cache Oblivious/Ext Memory/Parallel Delaunay

Page 10: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Biometrics

1. Recognition using hand outlines

2. One of the cheapest biometric solutions

3. Current Work: How to use 3D range scans of faces to do

face detection from images.

Page 11: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Other Problems

1. Detection of soil type using robot motion

2. Swarm robots (Covering type problems)

3. Cache oblivious and cache aware algorithms

4. Machine learning/Clustering.

5. Face recognition from 3d range scans.

Page 12: Research Overview Piyush Kumar. 1.Optimization 2.Computational Geometry 3.Computer Graphics 4.Pattern Recognition and Machine Learning / Biometrics 5.Robotics

Thanks for your attention

Questions?