Improved support vector machine generalization using normalized input space

Data pre-processing always plays a key role in learning algorithm performance. In this research we consider data pre-processing by normalization for Support Vector Machines (SVMs). We examine the normalization affect across 112 classification problems with SVM using the rbf kernel. We observe a significant classification improvement due to normalization. Finally we suggest a rule based method to find when normalization is necessary for a specific classification problem. The best normalization method is also automatically selected by SVM itself.

Simultaneous state and input estimation with application to a two-link robotic system

This paper addresses the problem of estimating simultaneously the state and input of a nonlinear system with application to a two link robotic manipulator - the Pendubot. The system nonlinearity comprises a Lipschitz function with respect to the state, and a nonlinear term which is a function of both the state and input. It is shown that under some conditions, an observer can be designed to estimate simultaneously the system’s state and input. Simulation and experimental results, obtained around the inverted equilibrium position, are presented to demonstrate the validity of the approach.

Design of reduced-order state/unknown input observers: a descriptor system approach

This paper addresses the problem of estimating simultaneously a linear function of both the state and unknown input of linear system with unknown inputs. By adopting the descriptor system approach, the problem can be conveniently solved. Observers proposed in this paper are of low-order and do not include the derivatives of the outputs. New conditions for the existence of reduced-order observers are derived. A design procedure for the determination of the observer parameters can also be easily derived based on the derived existence conditions

Informing decision-making in libraries: informetric research as input to LIS education and practice.(Library and Information Science).

Findings from informetric research represent an important background resource to add to the mix of information useful for resolving difficult and ongoing problems in specific library environments or information service settings. This paper provides examples of informetric research that can be useful input to decision-making in the field of library management and information service provision. This overview takes four of the challenges that Michael Buckland outlined for library research as a way of guiding the discussion of ways that informetric work can be used to inform library decision-making. (1) References are made to relevant informetric work undertaken or conducted in Australia, by Australian researchers, or with Australian data.

Informetrics includes both quantitative and qualitative methods, which when used in combination can provide a rounded set of findings that has great validity for management, policy and service applications. Quantitative methodologies are generally based on bibliometric techniques, such as mining and analysis of data from various bibliographic and textual databases. Qualitative methods include survey, case study and historical approaches. Used in combination, each set of findings adds richness and other perspectives to an analysis.

Detection and classification of road signs in natural environments

An automatic road sign recognition system first locates road signs within images captured by an imaging sensor on-board of a vehicle, and then identifies the detected road signs. This paper presents an automatic neural-network-based road sign recognition system. First, a study of the existing road sign recognition research is presented. In this study, the issues associated with automatic road sign recognition are described, the existing methods developed to tackle the road sign recognition problem are reviewed, and a comparison of the features of these methods is given. Second, the developed road sign recognition system is described. The system is capable of analysing live colour road scene images, detecting multiple road signs within each image, and classifying the type of road signs detected. The system consists of two modules: detection and classification. The detection module segments the input image in the hue-saturation-intensity colour space, and then detects road signs using a Multi-layer Perceptron neural-network. The classification module determines the type of detected road signs using a series of one to one architectural Multi-layer Perceptron neural networks. Two sets of classifiers are trained using the Resillient-Backpropagation and Scaled-Conjugate-Gradient algorithms. The two modules of the system are evaluated individually first. Then the system is tested as a whole. The experimental results demonstrate that the system is capable of achieving an average recognition hit-rate of 95.96% using the scaled-conjugate-gradient trained classifiers.

Using input-output data in life cycle inventory analysis

The impacts on the environment from human activities are of increasing concern. The need to consider the reduction in energy consumption is of particular interest, especially in the construction and operation of buildings, which accounts for between 30 and 40% of Australia's national energy consumption. Much past and more recent emphasis has been placed on methods for reducing the energy consumed in the operation of buildings. With the energy embodied in these buildings having been shown to account for an equally large proportion of a building's life cycle energy consumption, there is a need to look at ways of reducing the embodied energy of buildings and related products. Life cycle assessment (LCA) is considered to be the most appropriate tool for assessing the life cycle energy consumption of buildings and their products. The life cycle inventory analysis (LCIA) step of a LCA, where an inventory of material and energy inputs is gathered, may currently suffer from several limitations, mainly concerned with the use of incomplete and unreliable data sources and LCIA methods. These traditional methods of LCIA include process-based and input-output-based LCIA. Process-based LCIA uses process specific data, whilst input-output-based LCIA uses data produced from an analysis of the flow of goods and services between sectors of the Australian economy, also known as input-output data. With the incompleteness and unreliability of these two respective methods in mind, hybrid LCIA methods have been developed to minimise the errors associated with traditional LCIA methods, combining both process and input-output data. Hybrid LCIA methods based on process data have shown to be incomplete. Hybrid LCIA methods based on input-output data involve substituting available process data into the input-output model minimising the errors associated with process-based hybrid LCIA methods. However, until now, this LCIA method had not been tested for its level of completeness and reliability. The aim of this study was to assess the reliability and completeness of hybrid life cycle inventory analysis, as applied to the Australian construction industry. A range of case studies were selected in order to apply the input-output-based hybrid LCIA method and evaluate the subsequent results as obtained from each case study. These case studies included buildings: two commercial office buildings, two residential buildings, a recreational building; and building related products: a solar hot water system, a building integrated photovoltaic system and a washing machine. The range of building types and products selected assisted in testing the input-output-based hybrid LCIA method for its applicability across a wide range of product types. The input-output-based hybrid LCIA method was applied to each of the selected case studies in order to obtain their respective embodied energy results. These results were then evaluated with the use of a number of evaluation methods. These evaluation methods included an analysis of the difference between the process-based and input-output-based hybrid LCIA results as an evaluation of the completeness of the process-based LCIA method. The second method of evaluation used was a comparison between equivalent process and input-output values used in the input-output-based hybrid LCIA method as a measure of reliability. It was found that the results from a typical process-based LCIA and process-based hybrid LCIA have a large gap when compared to input-output-based hybrid LCIA results (up to 80%). This gap has shown that the currently available quantity of process data in Australia is insufficient. The comparison between equivalent process-based and input-output-based LCIA values showed that the input-output data does not provide a reliable representation of the equivalent process values, for material energy intensities, material inputs and whole products. Therefore, the use of input-output data to account for inadequate or missing process data is not reliable. However, as there is currently no other method for filling the gaps in traditional process-based LCIA, and as input-output data is considered to be more complete than process data, and the errors may be somewhat lower, using input-output data to fill the gaps in traditional process-based LCIA appears to be better than not using any data at all. The input-output-based hybrid LCIA method evaluated in this study has shown to be the most sophisticated and complete currently available LCIA method for assessing the environmental impacts associated with buildings and building related products. This finding is significant as the construction and operation of buildings accounts for a large proportion of national energy consumption. The use of the input-output-based hybrid LCIA method for products other than those related to the Australian construction industry may be appropriate, especially if the material inputs of the product being assessed are similar to those typically used in the construction industry. The input-output-based hybrid LCIA method has been used to correct some of the errors and limitations associated with previous LCIA methods, without the introduction of any new errors. Improvements in current input-output models are also needed, particularly to account for the inclusion of capital equipment inputs (i.e. the energy required to manufacture the machinery and other equipment used in the production of building materials, products etc.). Although further improvements in the quantity of currently available process data are also needed, this study has shown that with the current available embodied energy data for LCIA, the input-output-based hybrid LCIA appears to provide the most reliable and complete method for use in assessing the environmental impacts of the Australian construction industry.

Reforms in the Bangladesh agriculture input sector : room for complacency?

Market-based reforms implemented in the agricultural input management system are often claimed to be one of the success stories in Bangladesh. Based on the findings of 154 surveys and 16 in-depth interviews with the stakeholders during 2003-2004, the paper found that although there have been some notable achievements, the reforms have not achieved their expected outcomes and more importantly, in cases where they have been achieved, their sustainability is in question. The paper discusses the underlying reasons for the success and the problems of the reforms and also offers some policy recommendations to resolve the gap between the expected and the achieved outcomes. In the context of increasing donor pressure for adopting market-based reforms, the findings are likely to have important implications for both the donors and other countries in South Asia.

Simulation-based input loading condition optimisation of airport baggage handling systems

Scheduling check-in station operations are a challenging problem within airport systems. Prior to determining check-in resource schedules, an important step is to estimate the Baggage Handling System (BHS) operating capacity under non-stationary conditions. This ensures that check-in stations are not overloaded with bags, which would adversely affect the system and cause cascade stops and blockages. Cascading blockages can potentially lead to a poor level of service and in worst scenario a customer may depart without their bags. This paper presents an empirical study of a multiobjective problem within a BHS system. The goal is to estimate near optimal input operating conditions, such that no blockages occurs at check-in stations, while minimising the baggage travel time and maximising the throughput performance measures. We provide a practical hybrid simulation and binary search technique to determine a near optimal input throughput operating condition. The algorithm generates capacity constraint information that may be used by a scheduler to plan check-in operations based on flight arrival schedules.

An input-output-based linkage measurement framework

This study proposed an input-output-based linkage measurement framework, which is a multi-level hierarchy, omni-direction decision-making model and takes the impact of capital into account. This framework had resolved the critical deficiencies and inherent limitations of the existing methods and was used to explore the real estate and construction linkages.

Input and output analysis of complex conveyor networks

Material transfer networks are at the heart of critical infrastructure in many modern service and manufacturing industries. This research identified key performance measures, while deriving generalised analysis methodologies, for simulation models. The technology was validated for international airports, and used to determine operational capacity constraints under varied demand conditions.

Clinical pharmacy input for hospital inpatients according to disease state

This study measured the clinical activities performed and times taken by hospital pharmacists to provide medication monitoring services to individual medical and surgical patients. Linking these data to hospital Patient Administration Systems showed how clinical pharmacy manpower needs are guided by patient partition, disease complexity and Diagnosis Related Group classification.