Dennis Parkinson

Parallel efficiency can be greater than unity

Abstract It is shown that in some case parallel architectures with p processors can show speed-ups greater than p and efficiences greater than unity.

Autonomous Visual Events Detection and Classification without Explicit Object-Centred Segmentation and Tracking.

Modelling events is one of the key problems in dynamic scene analysis when salient and autonomous visual changes occuring in a scene need to be characterised effectively as meaningful events. We propose a new approach for modelling such temporal events based on the local intensity temporal history of pixels. The method provides a computationally very effective temporal measure for detecting autonomous events. Events are represented and detected first at the pixel level and then at a blob level (grouped pixels) autonomously. The Expectation-Maximisation (EM) algorithm is employed to cluster events with automatic model order selection using modified Minimum Description Length (MDL). Experiments are presented to demonstrate that meaningful clusters of blob-level events can be formed without object segmentation and tracking.

A compact algorithm for Gaussian elimination over GF(2) implemented on highly parallel computers

Gaussian elimination over GF(2) is used in a number of applications including the factorisation of large integers. The boolean nature of arithmetic in GF(2) makes the task well suited to highly parallel bit-organised computers. A program to work with up to 4096 x 4096 matrices has been developed for the ICL-DAP. A method has been developed that needs no extra storage to store the history of the elimination. The algorithm is presented and its correctness proved.

Algorithms for assignment problems on an array processor

Abstract The innovation of parallel computers has added a new dimension to the design of algorithms. Parallel programming is not a simple extension of serial programming. We describe parallel algorithms for the quadratic assignment problem and present our computational experience using the massively parallel processor, DAP. We further report the speedup obtained by parallelising algorithms for solving the 2-dimensional and 3-dimensional assignment problems on the DAP.

Optimum Tactics of Parallel Multi-Grid Algorithm with Virtual Boundary Forecast Method Running on a Local Network with the PVM Platform

In this paper, an optimum tactic of multi-grid parallel algorithm with virtual boundary forecast method is disscussed, and a two-stage implementation is presented. The numerical results of solving a non-linear heat transfer equation show that the optimum implementation is much better than the non-optimum one.

Outdoor Activity Recognition using Multi-Linked Temporal Processes

We develop Dynamically Multi-Linked Hidden Markov Models (DML-HMMs) for interpreting group activities involving multiple objects captured in an outdoor scene. The models are based on the discovery of salient dynamic interlinks among multiple different object events. A layered hierarchical DMLHMM is built using Schwarz’s Bayesian Information Criterion (BIC) based factorisation resulting in its topology being intrinsically determined by the underlying causality and temporal order among different object events. Our experiments demonstrate that the performance of a DML-HMM on modelling group activities in a noisy outdoor scene is superior compared to that of a Coupled Hidden Markov Model (CHMM).

Application of HPC to Medium-Size Stochastic Systems with Non-Linear Constraints in Finance

This paper presents results of multi-stage numerical optimisation of medium sized financial portfolios using High Performance Computing. Linear and non-linear constraints are applied and the optimisation process is designed to handle multiple markets in multiple countries, producing a series of alternative scenarios dependent upon a number of risk-rcturn requirements. The portfolio is allowed to contain instruments for which stochastic or historic data (or some combination) is available. Comparative performance is discussed for both real and artificial data sets and extrapolation to very large datasets will be presented. The comparative benefits of the deployment of large scale High Performance Computing in this class of problem are made.

Integrated presentation of 3-D data derived from multisensor imagery and anatomical atlases using a parallel processing system

The MIRIAD project (Medical Image Reconstruction, Interpretation, Analysis, and Display) brings together a multidisciplinary team with the objectives of exploring interactive presentation and model-based interpretation of three-dimensional medical images taken from high and low resolution studies respectively. A digitized atlas of normal anatomy can then be used to provide the personal atlas by which the medical image can be appraised and quantified.

Network computing performance evaluation in message passing environment

This paper is based on joint research between Queen Mary, University of London, and Wuhan University of Technology, Peoples Republic of China, on the use of network computing for solving non-linear heat flow problems. A general formulation for system performance evaluation has been derived from measured results, which covers the network computing as well as multiprocessor computing. The Amdahls law and Gustafsons modification for speedup have been uniformly explained from this formula. Some criterions of speedup, efficiency and granularity for network computing also have been suggested.