Object detection is a pervasive area of research in Computer Vision. Its applications are vast and include medical pathology (detection of malignant cells in histology slides), brain imaging (identification of structural signs of brain disease), detection of objects in remotely sensed imagery (traffic analysis and target detection), face detection for digital camera technology, and pedestrian tracking for surveillance. A broad spectrum of object detection approaches have been proposed in the literature, often using the popular boosting algorithm, AdaBoost. I propose a new object detection system that employs a new learning framework for combining ensembles of weak detectors into a single detector.

Prior work:

• Grammar-guided Feature Extraction from Signals and Images
  Image and signal sensors collect data at increasingly greater rates, making it a great challenge to search and organize data in archives. Work is underway to engage the sensor, data management, and machine learning expertise of LANL and UCSC to tackle adaptive, content-based search in large remote sensing archives. We demonstrate the utility of a new method for extracting features from imagery and signals to aid the archival search problem.
• CASCC: A Fast Algorithm for Time Series Classification
  The continual, real-time collection of time series signals makes manual analysis impractical demanding faster and more accurate analysis algorithms.  We are developing algorithms for rapid clustering and classification of time series data sets.  CASCC is a new algorithm for classifying time series. It is highly competitive in terms of speed and accuracy compared to many other algorithms. It is inspired by another leading alghorthm DTW-1NN howerver does not suffer the same computational limitations when applying the model to new time series.

Damian Eads
UCSC Student

Ethan Miller
UCSC Instructor

Hai Tao
UCSC Instructor

David Helmbold
UCSC Instructor

James Theiler
Mentor

Simon Perkins
Mentor

Edward Rosten
Mentor

Andrew Fraser
Mentor

• Learning Object Detectors for Panchromatic Imagery
• Grammar-guided Feature Extraction from Signals and Images Poster
• Advances and Defenses: Ensembles of Weak Object Detectors
• CASCC: A Fast Algorithm for Time Series Classification Poster
• paper
• paper
• Paper in process
• British Machine Vision Conference Paper

Research Instruments Used:

• ISSDM Cluster
• Googoo Cluster

RAID systems have traditionally offered increased performance and data security in small storage systems. An opportunity now exists to extend traditional RAID principles into the area of large-scale object-based storage devices in order to offer greater data security and space efficiency. In a system where component failures can be expected on a daily basis, the importance of redundancy mechanisms is obvious, and RAID principles offer an appropriate model. Ceph is an excellent platform to test these RAID systems and to learn how they function in a new environment.

David Bigelow
UCSC Student

Scott Brandt
UCSC Instructor

Gary Grider
Mentor

James Nunez
Mentor

John Bent
Mentor

HB Chen
Mentor

• Alternative Reliability Models in Ceph Poster
• ISSDM Day 2009 Presentation

The ISIS team in ISR-2 primarily uses supervised learning techniques to solve classification problems in imagery and therefore has a strong interest in finding linear classification algorithms that are both robust and efficient. Boosting algorithms take a principled approach to finding linear classifiers and they has been shown to be so effective in practice that they are widely used in a variety of domains. In this proposal we present evidence that smoothing is not necessarily the optimal way

Karen Glocer
UCSC Student

Manfred Warmuth
UCSC Instructor

James Theiler
Mentor

Simon Perkins
Mentor

• Entropy Regularized LPBoost Alt 08 talk
• Boosting for Image Recognition Proposal
• Talk
• Talk about paper
• paper
• paper

Research Instruments Used:

• ISSDM Cluster

Tuning is one of the most challenging and important tasks in setting up a database system. A typical assumption is that a representative workload can provide the means to perform some initial tuning automatically, which can then be refined manually by the database administrator. However, several emerging application domains challenge this assumption. A characteristic example is scientific data
management, where the use of ad-hoc data analytics cancels the ability to gather a representative workload, and the lack of database expertise among scientists lowers the possibility for efficient manual tuning. This motivates us to look into autonomic tuning techniques that operate continuously and require minimal intervention from an administrator. In particular, we examine online methods to tune the physical database design, i.e., the set of materialized data structures, such as indices and materialized views, that are crucial for efficient query processing.

Karl Schnaitter
UCSC Student

Neoklis Polyzotis
UCSC Instructor

Gary Grider
Mentor

James Nunez
Mentor

John Bent
Mentor

• COLT: Continuous On-Line Tuning of Databases Proposal
• Thesis Proposal
• VLDB 2009 Paper
• SMDB 2007 Workshop Paper
• UCSC Technical Report, 2006
• SMDB 2009 Workshop Paper
• UCSC Technical Report, 2009
• SIGMOD 2006 Demo
• ISSDM Day 2009 Presentation

Research Instruments Used:

• Googoo Cluster

We propose to work on the problem of calibrating the parameters of computer code used for simulation of physical phenomena. We will explore statistical methods based on a Bayesian approach implemented with Sampling Importance Resampling (SIR).

Tracy Holsclaw
UCSC Student

Herbie Lee
UCSC Instructor

Bruno Sanso
UCSC Instructor

David Higdon
Mentor

Katrin Heitmann
Mentor

Salman Habib
Mentor

• Cosmic Calibration Presentation
• Cosmic Calibration poster at UCSC Engr Day
• Statistical Models for Dark Energy
• Cosmic Calibration Poster-Statistical Modeling for Dark Energy
• Cosmic Calibration Proposal
• ISSDM Day 2009 Paper
• ISSDM Day 2009 Poster
• ISSDM Day 2009 Presentation

Pseudorandom placement in distributed storage systems offers scalability benefits. Pseudorandom placement makes load balancing harder; new techniques are required. We explore different load balancing techniques using Ceph, an object-based storage system developed at UCSC.

Esteban Molina-Estolano
UCSC Student

Scott Brandt
UCSC Instructor

Gary Grider
Mentor

James Nunez
Mentor

John Bent
Mentor

• Dynamic Load-Balancing File Systems FAST 09 WIP Talk
• Dynamic Load-Balancing in Petabyte-scale File Systems that use Pseudo-Random Data
• Dynamic Load-Balancing Presentation
• Abstract for SciDAC 2009
• ISSDM Day 2009 Presentation

Large data centers are typically composed of many hard drives arranged in a striped layout with redundancy, to provide adequate performance and reliability. Solid state disk drives (SSDs) are a newer technology that has better performance and lower power requirements compared to hard drives. This research examines the use of solid state disk drives in disk-SSD hybrid storage systems.

Rosie Wacha
UCSC Student

Scott Brandt
UCSC Instructor

Gary Grider
Mentor

James Nunez
Mentor

John Bent
Mentor

• Erasure Codes for Reliability and Recoverability Abstract
• Ph.D. Advancement Proposal
• hotemnets 08 paper
• Disaster Recovery Codes: Increasing Reliability with Large-Stripe Erasure Correcting Codes
• Self-Adaptive Two-Dimensional RAID Arrays
• UCSC Graduate Research Symposium Poster
• ISSDM Day 2009 Presentation

Definition: Mutivalue data are where you have multiple values about the same variable in repeated measurements.

Challenge: Lack of current comprehensive visualization tools for the probabilistic or uncertain nature of the multivalue data.

Research: Investigate probabilistic streamlines using different ways of combining vector distributions.

The size of the data sets and the uncertainty in the data sets come from the fact that we are dealing with ensemble data sets. These are usually from Monte Carlo simulations where each output (out of many runs) represents a possible solution. The degree of agreement (or disagreement) provides some indications of certainty (or uncertainty) about the results. Because Monte Carlo simulations can potentially involve large number of repetitions, the total data size can very quickly get very large. This project will explore uncertainty and how visualization can be used as a tool to help deal with it.

Eddy Chandra
UCSC Student

Alex Pang
UCSC Instructor

Katrin Heitmann
Mentor

James Ahrens
Mentor

• Uncertainty Visualization Poster
• Project Interim Results
• Exploring Uncertainty Visualization in Large Data Sets Propopsal
• Paper
• ISSDM Day 2009 Presentation

To help users make sense of collaboratively-generated information, we are developing algorithmic notions of information trust

Ian Pye
UCSC Student

Bo Adler
UCSC Student

Luca de Alfaro
UCSC Instructor

Shelly Spearing
Mentor

Jorge Roman
Mentor

• Information Trust Proposal
• paper
• paper
• paper
• paper
• ISSDM Day 2009 Presentation - SCRAWL
• ISSDM Day 2009 Presentation - WikiTrust

Research Instruments Used:

• ISSDM Cluster

We focus on developing intelligent interactive search and browsing techniques to help users and the information they are looking for from billions of non-relevant files

Jessica Gronski
UCSC Student

Yi Zhang
UCSC Instructor

Herbert Van de Sompel
Mentor

• Interactive Search and Browsing Proposal

In large tightly coupled parallel systems, computation goes as fast as the slowest part.  For this reason it is necessary to pursue deterministic behavior of all parts of the system.  Quality of Service is one way to assist in providing deterministic behavior.  This project will explore providing Quality of Service on networks of interest to high performance computing.

Andrew Shewmaker
UCSC Student

Scott Brandt
UCSC Instructor

Andrew Shewmaker
Mentor

• Network Quality of Service SSRC 08 Poster
• ISSDM Day 2009 Presentation

The research objective of this proposal is to measure human body shape and motion without augmenting the subject. The hypothesis is that replacing traditional cameras with high accuracy 3D shape measurement devices and utilizing a carefully constructed prior model of human surface shape are the critical factors that have been missing from prior attempts to meet this goal. The long term accuracy targets are shape to 1mm and motion to 1deg.

Steve Scher
UCSC Student

James Davis
UCSC Instructor

Sriram Swaminarayan
Mentor

• Sensing Human Shape and Motion Proposal
• International Conference on Pattern Recognition Poster
• Paper
• conference web site
• conference url
• ISSDM Day 2009 Presentation

We are planning to develop a web-based system to help 'decision makers'
quickly identify and process relevant web-based information in case of a
disease outbreak. We will work on identifying the pathogen based on sequence
information. We will also develop an adaptive information filtering to find,
filter and condense the information available on the web.

Yi Zhang
UCSC Instructor

Carla Kuiken
Mentor