Search

This paper describes improvements to the rule chaining architecture presented in [1]. The architecture uses distributed associative memories to allow the system to utilise memory efficiently, and superimposed distributed representations in order to reduce the time complexity of a tree search to O(d), where d is the depth of the tree. This new work reduces the memory required by the architecture, and can also further reduce the time complexity.

This paper describes an architecture based on superimposed distributed representations and distributed associative memories which is capable of performing rule chaining. The use of a distributed representation allows the system to utilise memory efficiently, and the use of superposition reduces the time complexity of a tree search to O(d), where d is the depth of the tree. Our experimental results show that the architecture is capable of rule chaining effectively, but that further investigation is needed to address capacity considerations.

This chapter presents the Relaxation by Elimination methods (RBE) for searching large collections of graph data that has been implemented on a distributed platform and is in daily use for searching a database of molecules. The core of the approach uses an ‘inverted’ relaxation labelling method that finds a good match of the input data with stored examples. The method is shown to scale linearly with the number of graphs, and to scale linearly under given circumstances to the number of nodes in the graph. Key to the idea is that the system cuts the search time by removing a set of sub-optimal matches leaving those that could match. The system uses arrays of biologically plausible neural networks, Correlation Matrix Memories (CMMs) to store the constraints between the nodes of the graphs being searched. This is coupled to a novel search method. The system is highly parallel. Recently we have developed a parallel Grid enabled computer system (Cortex II) which utilises Digital Signal Processors (DSPs) and Field Programmable Gate Arrays (FPGAs) and have implemented the method on this system. A service for matching small molecules to a molecule database, in which the molecules are represented as attributed graphs, is currently running online. The methods have also been applied to searching trademark databases allowing people to find trademarks that are geometrically similar. The chapter describes the method in detail and its implementation and application. It also brings together work that has appeared separately and presents a new mathematical formulation of the mapping of RBE onto correlation matrix methods.

We describe our approach to segmenting moving objects from the color video data supplied by a nominally stationary camera. There are two main contributions in our work. The first contribution augments Zivkovic and Heijden's recursively updated Gaussian mixture model approach, with a multi-dimensional Gaussian kernel spatio-temporal smoothing transform. We show that this improves the segmentation performance of the original approach, particularly in adverse imaging conditions, such as when there is camera vibration. Our second contribution is to present a comprehensive comparative evaluation of shadow and highlight detection appoaches, which is an essential component of background subtraction in unconstrained outdoor scenes. A comparative evelaution of these approaches over different color-spaces is currently lacking in the literature. We show that both segmentation and shadow removal performs best when we use RGB color spaces.

In this paper we introduce an improved binary correlation matrix memory (CMM) with better storage capacity when storing sparse fixed weight codes generated with the algorithm of Baum et al. [3]. We outline associative memory, and describe the binary correlation matrix memory-- a specific example of a distributed associative memory. The importance of the representation used in a CMM for input and output codes is discussed, with specific regard to sparse fixed weight codes. We present an algorithm for generating of fixed weight codes, originally given by Baum et al. [3]. The properties of this algorithm are briefly discussed, including possible thresholding functions which could be used when storing these codes in a CMM; L-max and L-wta. Finally, results generated from a series of simulations are used to demonstrate that the use of L-wta as a thresholding function provides an increase in storage capacity over L-max.

In this paper, we investigate human visual perception and establish a body of ground truth data elicited from human visual studies. We aim to build on the formative work of Ren, Eakins and Briggs who produced an initial ground truth database. Human participants were asked to draw and rank their perceptions of the parts of a series of figurative images. These rankings were then used to score the perceptions, identify the preferred human breakdowns and thus allow us to induce perceptual rules for human decomposition of figurative images. The results suggest that the human breakdowns follow well-known perceptual principles in particular the Gestalt laws.

This paper defines a constrained Artificial Neural Network (ANN) that can be employed for highly-dependable roles in safety critical applications. The derived model is based upon the Fuzzy Self-Organising Map (FSOM) and enables behaviour to be described qualitatively and quantitatively. By harnessing these desirable features, behaviour is bounded through incorporation of safety constraints – derived from safety requirements and hazard analysis. The constrained FSOM has been termed a ’Safety Critical Artificial Neural Network’ (SCANN) and preserves valuable performance characteristics for non-linear function approximation problems. The SCANN enables construction of compelling (product-based) safety arguments for mitigation and control of identified failure modes. Illustrations of potential benefits for real-world applications are also presented.

AURA (Advanced Uncertain Reasoning Architecture) is a parallel pattern matching technology intended for high-speed approximate search and match operations on large unstructured datasets. This paper represents how the AURA technology is extended and used to search the engine data within a major UK eScience Grid project (DAME) for maintenance of Rolls-Royce aero-engines and how it may be applied in other areas. Examples of its use will be presented.

The k-Nearest Neighbour (kNN) approach is a widely-used technique for pattern classification. Ranked distance measurements to a known sample set determine the classification of unknown samples. Though effective, kNN, like most classification methods does not scale well with increased sample size. This is due to their being a relationship between the unknown query and every other sample in the data space. In order to make this operation scalable, we apply AURA to the kNN problem. AURA is a highly-scalable associative-memory based binary neural-network intended for high-speed approximate search and match operations on large unstructured datasets. Previous work has seen AURA methods applied to this problem as a scalable, but approximate kNN classifier. This paper continues this work by using AURA in conjunction with kernel-based input vectors, in order to create a fast scalable kNN classifier, whilst improving recall accuracy to levels similar to standard kNN implementations.

Present approaches for computing do not have the performance, flexibility and reliability of neural information processing systems. In order to overcome this, conventional computing systems could benefit from various characteristics of the brain such as modular organisation, robustness, timing and synchronisation, and learning and memory storage in the central nervous system. This overview incorporates some of the key research issues in the field of biologically inspired computing systems.

This book is the result of a series of International Workshops organised bythe EmerNet project on Emergent Neural Computational Architectures basedon Neuroscience sponsored by the Engineering and Physical Sciences ResearchCouncil (EPSRC). The overall aim of the book is to present a broad spectrum ofcurrent research into biologically inspired computational systems and hence en-courage the emergence of new computational approaches based on neuroscience.It is generally understood that the present approaches for computing do not havethe performance, exibility and reliability of biological information processingsystems. Although there is a massive body of knowledge regarding how process-ing occurs in the brain and central nervous system this has had little impact onmainstream computing so far.The process of developing biologically inspired computerised systems involvesthe examination of the functionality and architecture of the brain with an empha-sis on the information processing activities. Biologically inspired computerisedsystems address neural computation from the position of both neuroscience,and computing by using experimental evidence to create general neuroscience-inspired systems.The book focuses on the main research areas of modular organisation androbustness, timing and synchronisation, and learning and memory storage. Theissues considered as part of these include: How can the modularity in the brainbe used to produce large scale computational architectures? How does the hu-man memory manage to continue to operate despite failure of its components?How does the brain synchronise its processing? How does the brain computewith relatively slow computing elements but still achieve rapid and real-timeperformance? How can we build computational models of these processes andarchitectures? How can we design incremental learning algorithms and dynamicmemory architectures? How can the natural information processing systems beexploited for arti cial computational methods?We hope that this book stimulates and encourages new research in this area.We would like to thank all contributors to this book and the few hundred partici-pants of the various workshops. Especially we would like to express our thanks toMark Elshaw, network assistant in the EmerNet network who put in tremendouse ort during the process of publishing this book.Finally, we would like to thank EPSRC and James Fleming for their supportand Alfred Hofmann and his sta at Springer for their continuing assistance.March 2001Stefan WermterJim AustinDavid Willshaw

A novel purely connectionist implementation of propositional logic is constructed by combining Correlation Matrix Memory operations, tensor products and simple control circuits. The implementation is highly modular and expandable and in its present form it not only allows forward rule chaining but also implements is a hierarchy traversal which results in interesting behavior even in its simplest form.

In this paper we propose a simple, flexible and efficient hybrid spell checking methodology based upon phonetic matching, supervised learning and associative matching in the AURA neural system. We evaluate our approach against several benchmark spell-checking algorithms for recall accuracy. Our proposed hybrid methodology has the joint highest top 10 recall rate of the techniques evaluated. The method has a high recall rate and low computational cost.

The aim of the research reported in this paper was to assess the scalability of a binary correlation Matrix Memory (CMM) based on the PRESENCE (PaRallEl StructurEd Neural Computing Engine) architecture. A single PRESENCE card has a finite memory capacity, and this paper describes howm ultiple PCI-based PRESENCE cards are utilised in order to scale up memory capacity and performance. A Beowulf class cluster, called Cortex-1, provides the scalable I/O capacity needed for multiple cards, and techniques for mapping applications onto the system are described. The main aims of the work are to prove the scalability of the AURA architecture, and to demonstrate the capabilities of the architecture for commercial pattern matching problems.

A system which combines the descriptional power of symbolic representations with the parallel and distributed processing model of cellular automata and the speed and robustness of connectionist symbol processing is described. Following a cellular automata based approach, the aim of the system is to transform initial symbolic descriptions of patterns to corresponding object level descriptions in order to identify patterns in complex or noisy scenes. A learning algorithm based on a hierarchical structural analysis is used to learn symbolic descriptions of objects. The underlying symbolic processing engine of the system is a neural based associative memory (AURA) which enables the system to operate in high speed. In addition, the use of distributed representations allow both efficient inter-cellular communications and compact storage of rules.

This paper describes a novel massively parallel connectionist architecture for image retrieval. The proposed search engine of the system consists of associative memory nodes connected by information channels which convey symbolic messages. Symbolic information stored inside the system is obtained using gestalt feature extraction methods which capture multiple representations of images. In this paper, we summarise our feature extraction method and then we describe the connection schemata of the system, training process as well as how such a system can be utilised to capture perceptual similarity of trademark images. Finally, we present results obtained during evaluation of the system.

In the very near future large amounts of Remotely Sensed data will become available on a daily basis. Unfortunately, it is not clear if the processing methods are available to deal with this data in a timely fashion. This paper describes research towards an approach which will allow a user to perform a rapid pre-search of large amounts of image data for regions of interest based on texture. The method is based on a novel neural network architecture (ADAM) that is designed primarily for speed of operation by making use of computationally simple pre-processing and only uses Boolean operations in the weights of the network. To facilitate interactive use of the network, it is capable of rapid training. The paper outlines the neural network, its application to RS data in comparison with other methods, and briefly describes a fast hardware implementation of the network.

The European Concerted Action “COMPARES” (Concerted Action on COnnectionist Methods for Preprocessing and Analysis of REmote Sensing Data) was funded within the Environment and Climate Programme of the European Commission. The “COMPARES” project was aimed at focussing and coordinating future research at European level on the general theme of connectionist computation in the field of Earth Observation (EO). Connectionist computation can be interpreted as the use of artificial neural networks and parallel distributed processing systems.

This paper describes a generalisation of the binary N tuple technique originally described by Bledsoe and Browning (1). The binary N tuple technique has commonly been used for the classification (2) and pre-processing (3) of binary images. The extension to the method described here allows grey level images of objects to be classified using the N tuple method without first having to convert the image to an intermediate binary representation. The paper illustrates the methods use in image preprocessing.