GYAN: A methodology for rule extraction from artificial neural networks

Artificial neural network (ANN) learning methods provide a robust and non-linear approach to approximating the target function for many classification, regression and clustering problems. ANNs have demonstrated good predictive performance in a wide variety of practical problems. However, there are strong arguments as to why ANNs are not sufficient for the general representation of knowledge. The arguments are the poor comprehensibility of the learned ANN, and the inability to represent explanation structures. The overall objective of this thesis is to address these issues by: (1) explanation of the decision process in ANNs in the form of symbolic rules (predicate rules with variables); and (2) provision of explanatory capability by mapping the general conceptual knowledge that is learned by the neural networks into a knowledge base to be used in a rule-based reasoning system. A multi-stage methodology GYAN is developed and evaluated for the task of extracting knowledge from the trained ANNs. The extracted knowledge is represented in the form of restricted first-order logic rules, and subsequently allows user interaction by interfacing with a knowledge based reasoner. The performance of GYAN is demonstrated using a number of real world and artificial data sets. The empirical results demonstrate that: (1) an equivalent symbolic interpretation is derived describing the overall behaviour of the ANN with high accuracy and fidelity, and (2) a concise explanation is given (in terms of rules, facts and predicates activated in a reasoning episode) as to why a particular instance is being classified into a certain category.

Multiple camera management using wide baseline matching

Camera calibration information is required in order for multiple camera networks to deliver more than the sum of many single camera systems. Methods exist for manually calibrating cameras with high accuracy. Manually calibrating networks with many cameras is, however, time consuming, expensive and impractical for networks that undergo frequent change. For this reason, automatic calibration techniques have been vigorously researched in recent years. Fully automatic calibration methods depend on the ability to automatically find point correspondences between overlapping views. In typical camera networks, cameras are placed far apart to maximise coverage. This is referred to as a wide base-line scenario. Finding sufficient correspondences for camera calibration in wide base-line scenarios presents a significant challenge. This thesis focuses on developing more effective and efficient techniques for finding correspondences in uncalibrated, wide baseline, multiple-camera scenarios. The project consists of two major areas of work. The first is the development of more effective and efficient view covariant local feature extractors. The second area involves finding methods to extract scene information using the information contained in a limited set of matched affine features. Several novel affine adaptation techniques for salient features have been developed. A method is presented for efficiently computing the discrete scale space primal sketch of local image features. A scale selection method was implemented that makes use of the primal sketch. The primal sketch-based scale selection method has several advantages over the existing methods. It allows greater freedom in how the scale space is sampled, enables more accurate scale selection, is more effective at combining different functions for spatial position and scale selection, and leads to greater computational efficiency. Existing affine adaptation methods make use of the second moment matrix to estimate the local affine shape of local image features. In this thesis, it is shown that the Hessian matrix can be used in a similar way to estimate local feature shape. The Hessian matrix is effective for estimating the shape of blob-like structures, but is less effective for corner structures. It is simpler to compute than the second moment matrix, leading to a significant reduction in computational cost. A wide baseline dense correspondence extraction system, called WiDense, is presented in this thesis. It allows the extraction of large numbers of additional accurate correspondences, given only a few initial putative correspondences. It consists of the following algorithms: An affine region alignment algorithm that ensures accurate alignment between matched features; A method for extracting more matches in the vicinity of a matched pair of affine features, using the alignment information contained in the match; An algorithm for extracting large numbers of highly accurate point correspondences from an aligned pair of feature regions. Experiments show that the correspondences generated by the WiDense system improves the success rate of computing the epipolar geometry of very widely separated views. This new method is successful in many cases where the features produced by the best wide baseline matching algorithms are insufficient for computing the scene geometry.

Assessment of the suitability of public mobile data networks for aircraft telemetry and control purposes

This paper provides a review of the state of the art relevant work on the use of public mobile data networks for aircraft telemetry and control proposes. Moreover, it describes the characterisation for airborne uses of the public mobile data communication systems known broadly as 3G. The motivation for this study was the explore how this mature public communication systems could be used for aviation purposes. An experimental system was fitted to a light aircraft to record communication latency, line speed, RF level, packet loss and cell tower identifier. Communications was established using internet protocols and connection was made to a local server. The aircraft was flown in both remote and populous areas at altitudes up to 8500 ft in a region located in South East Queensland, Australia. Results show that the average airborne RF levels are better than those on the ground by 21% and in the order of - 77dbm. Latencies were in the order of 500ms (1/2 the latency of Iridium), an average download speed of 0.48Mb/s, average uplink speed of 0.85Mb/s, a packet of information loss of 6.5%. The maximum communication range was also observed to be 70km from a single cell station. The paper also describes possible limitations and utility of using such communications architecture for both manned and unmanned aircraft systems.

Camera calibration in wireless multimedia sensor networks

Wireless Multi-media Sensor Networks (WMSNs) have become increasingly popular in recent years, driven in part by the increasing commoditization of small, low-cost CMOS sensors. As such, the challenge of automatically calibrating these types of cameras nodes has become an important research problem, especially for the case when a large quantity of these type of devices are deployed. This paper presents a method for automatically calibrating a wireless camera node with the ability to rotate around one axis. The method involves capturing images as the camera is rotated and computing the homographies between the images. The camera parameters, including focal length, principal point and the angle and axis of rotation can then recovered from two or more homographies. The homography computation algorithm is designed to deal with the limited resources of the wireless sensor and to minimize energy con- sumption. In this paper, a modified RANdom SAmple Consensus (RANSAC) algorithm is proposed to effectively increase the efficiency and reliability of the calibration procedure.

Network structure, knowledge governance, and firm performance : evidence from innovation networks and SMEs in the UK

It is increasingly understood that learning and thus innovation often occurs via highly interactive, iterative, network-based processes. Simultaneously, economic development policy is increasingly focused on small and medium-sized enterprises (SMEs) as a means of generating growth, creating a clear research issue in terms of the roles and interactions of government policy, universities, and other sources of knowledge, SMEs, and the creation and dissemination of innovation. This paper analyses the contribution of a range of actors in an SME innovation creation and dissemination framework, reviewing the role of various institutions therein, exploring the contribution of cross-locality networks, and identifying the mechanisms required to operationalise such a framework. Bivariate and multivariate (regression) techniques are employed to investigate both innovation and growth outcomes in relation to these structures; data are derived from the survey responses of over 450 SMEs in the UK. Results are complex and dependent upon the nature of institutions involved, the type of knowledge sought, and the spatial level of the linkages in place but overall highlight the value of cross-locality networks, network governance structures, and certain spillover effects from universities. In general, we find less support for the factors predicting SME growth outcomes than is the case for innovation. Finally, we outline an agenda for further research in the area.

Detecting attacks in encrypted networks using secret-sharing schemes

Secret-sharing schemes describe methods to securely share a secret among a group of participants. A properly constructed secret-sharing scheme guarantees that the share belonging to one participant does not reveal anything about the shares of others or even the secret itself. Besides being used to distribute a secret, secret-sharing schemes have also been used in secure multi-party computations and redundant residue number systems for error correction codes. In this paper, we propose that the secret-sharing scheme be used as a primitive in a Network-based Intrusion Detection System (NIDS) to detect attacks in encrypted Networks. Encrypted networks such as Virtual Private Networks (VPNs) fully encrypt network traffic which can include both malicious and non-malicious traffic. Traditional NIDS cannot monitor such encrypted traffic. We therefore describe how our work uses a combination of Shamir's secret-sharing scheme and randomised network proxies to enable a traditional NIDS to function normally in a VPN environment.

Photo-cross-linked poly(D,L-lactide)-based networks. Structural characterization by HR-MAS NMR spectroscopy and hydrolytic degradation behavior

To date, biodegradable networks and particularly their kinetic chain lengths have been characterized by analysis of their degradation products in solution. We characterize the network itself by NMR analysis in the solvent-swollen state under magic angle spinning conditions. The networks were prepared by photoinitiated cross-linking of poly(dl-lactide)−dimethacrylate macromers (5 kg/mol) in the presence of an unreactive diluent. Using diffusion filtering and 2D correlation spectroscopy techniques, all network components are identified. By quantification of network-bound photoinitiator fragments, an average kinetic chain length of 9 ± 2 methacrylate units is determined. The PDLLA macromer solution was also used with a dye to prepare computer-designed structures by stereolithography. For these networks structures, the average kinetic chain length is 24 ± 4 methacrylate units. In all cases the calculated molecular weights of the polymethacrylate chains after degradation are maximally 8.8 kg/mol, which is far below the threshold for renal clearance. Upon incubation in phosphate buffered saline at 37 °C, the networks show a similar mass loss profile in time as linear high-molecular-weight PDLLA (HMW PDLLA). The mechanical properties are preserved longer for the PDLLA networks than for HMW PDLLA. The initial tensile strength of 47 ± 2 MPa does not decrease significantly for the first 15 weeks, while HMW PDLLA lost 85 ± 5% of its strength within 5 weeks. The physical properties, kinetic chain length, and degradation profile of these photo-cross-linked PDLLA networks make them most suited materials for orthopedic applications and use in (bone) tissue engineering.

Project-based organisations : overcoming lack of trust and social networks within small and medium project teams

In sustainable development projects, as well as other types of projects, knowledge transfer is important for the organisations managing the project. Nevertheless, knowledge transfer among employees does not happen automatically and it has been found that the lack of social networks and the lack of trust among employees are the major barriers to effective knowledge transfer. Social network analysis has been recognised as a very important tool for improving knowledge transfer in the project environment. Transfer of knowledge is more effective where it depends heavily on social networks and informal dialogue. Based on the theory of social capital, social capital consists of two parts: conduits network and resource exchange network. This research studies the relationships among performance, the resource exchange network (such as the knowledge network) and the relationship network (such as strong ties network, energy network, and trust network) at the individual and project levels. The aim of this chapter is to present an approach to overcoming the lack of social networks and lack of trust to improve knowledge transfer within project-based organisations. This is to be done by identifying the optimum structure of relationship networks and knowledge networks within small and medium projects. The optimal structure of the relationship networks and knowledge networks is measured using two dimensions: intra-project and inter-project. This chapter also outlines an extensive literature review in the areas of social capital, knowledge management and project management, and presents the conceptual model of the research approach.

Impact of embeddedness in co-ethnic & non co-ethnic networks on business performance : evidence from Chinese immigrant entrepreneurs in Australia

This paper investigates the research question ‘What is the effect of co-ethnic and non coethnic networking on business performance in Chinese immigrant businesses?’ The research will discuss key themes such as the extent to which Chinese immigrant entrepreneurs are embedded in co-ethnic and non co-ethnic networks and the affect of embeddedness on business performance, such as the entrepreneur’s satisfaction and business growth. Research on immigrant entrepreneurship has emerged as an important new area of inquiry within the field of entrepreneurship. The increased importance of the subject is due in part to major immigrant receiving countries, such as Australia, the United States and Canada, experiencing a high growth rate in their immigrant population. Reflecting on the existing research on immigrant entrepreneurship, it was decided to investigate the role of embeddedness on entrepreneurial business performance. This research seeks to identify the impact of embeddedness in co-ethnic and non co-ethnic networks on business performance of Chinese immigrant entrepreneurs in Australia. Chinese immigrant restaurant entrepreneurs in southeast Queensland, Australia were studied. The result expands on existing research on immigrant entrepreneurship, since the majority of immigrant entrepreneurship studies have been conducted on the United States and Canada immigrant experiences, but few have been conducted in the Australian immigrant entrepreneur context. This thesis also adds empirical testing to a research area with little empirical testing. The results indicated that embeddedness in the co-ethnic network is positively related to business performance measured by both growth and satisfaction. Embeddedness in the non co-ethnic network of the Chinese immigrant entrepreneurs in Australia did not show a similar pattern in accordance with studies conducted in the United States and Canada. This result is interesting and creates the opportunity for future research employing a comparative study.

A distributed protocol for object tracking in wireless multimedia sensor networks

We consider the problem of object tracking in a wireless multimedia sensor network (we mainly focus on the camera component in this work). The vast majority of current object tracking techniques, either centralised or distributed, assume unlimited energy, meaning these techniques don't translate well when applied within the constraints of low-power distributed systems. In this paper we develop and analyse a highly-scalable, distributed strategy to object tracking in wireless camera networks with limited resources. In the proposed system, cameras transmit descriptions of objects to a subset of neighbours, determined using a predictive forwarding strategy. The received descriptions are then matched at the next camera on the objects path using a probability maximisation process with locally generated descriptions. We show, via simulation, that our predictive forwarding and probabilistic matching strategy can significantly reduce the number of object-misses, ID-switches and ID-losses; it can also reduce the number of required transmissions over a simple broadcast scenario by up to 67%. We show that our system performs well under realistic assumptions about matching objects appearance using colour.

Decentralised particle filtering for multiple target tracking in wireless sensor networks

Record 8 of 29