In recent years, various metrics have been developed for measuring the similarity of states in probabilistic transition systems (Desharnais et al., 1999; van Breugel & Worrell, 2001a). In the context of Markov decision processes, we have devised metrics providing a robust quantitative analogue of bisimulation. Most importantly, the metric distances can be used to bound the differences in the optimal value function that is integral to reinforcement learning (Ferns et al. 2004; 2005). More recently, we have discovered an efficient algorithm to calculate distances in the case of finite systems (Ferns et al., 2006). In this thesis, we seek to properly extend state-similarity metrics to Markov decision processes with continuous state spaces both in theory and in practice. In particular, we provide the first distance-estimation scheme for metrics based on bisimulation for continuous probabilistic transition systems. Our work, based on statistical sampling and infinite dimensional linear programming, is a crucial first step in real-world planning; many practical problems are continuous in nature, e.g. robot navigation, and often a parametric model or crude finite approximation does not suffice. State-similarity metrics allow us to reason about the quality of replacing one model with another. In practice, they can be used directly to aggregate states.

No description available.

No description available.

There is no abstract available for this thesis.

The duration of the post reinforcement pause was measured for three white rats on fixed ratio schedules of reinforcement ranging from 26 to 120. The post reinforcement pause consistently increased as the ratio requirement increased for all three subjects. The data failed to show one function (power, logarithmic, or linear) as significantly more useful than the other two in describing the relationship between systematically increased fixed ratios and the respective post reinforcement pause for each ratio. Furthermore, there was no consistent relationship between the response rate and the fixed ratio requirement.

The relationship between level of deprivation and post-reinforcement pause length was investigated in such a way that the nature of the mathematical function which describes the relation could be determined. Three male, albino rats were tested on fixed-ratio schedules of reinforcement at three levels of deprivation, according to the method of adjusted-percentages. Problems with the weight maintenance procedure necessitated changes in the design and may have affected the results. The initial hypothesis, that the function would be of the form Y=bXn, was not supported in any case. A function of the form logY=a+bX was found to best describe the relationship between rats' absolute body weight and the length of the post-reinforcement pause. Conclusions were centered around the need for further research on the effects of long-term deprivation on rats' growth and subsequent reexamination of the relationship under investigation in this thesis.

No description available.

In this thesis the development of a hardware artificial neuron device and artificial neural network using stochastic pulse rate encoding principles is considered. After a review of neural network architectures and algorithmic approaches suitable for hardware implementation, a critical review of hardware techniques which have been considered in analogue and digital systems is presented. New results are presented demonstrating the potential of two learning schemes which adapt by the use of a single reinforcement signal. The techniques for computation using stochastic pulse rate encoding are presented and extended with new novel circuits relevant to the hardware implementation of an artificial neural network. The generation of random numbers is the key to the encoding of data into the stochastic pulse rate domain. The formation of random numbers and multiple random bit sequences from a single PRBS generator have been investigated. Two techniques, Simulated Annealing and Genetic Algorithms, have been applied successfully to the problem of optimising the configuration of a PRBS random number generator for the formation of multiple random bit sequences and hence random numbers. A complete hardware design for an artificial neuron using stochastic pulse rate encoded signals has been described, designed, simulated, fabricated and tested before configuration of the device into a network to perform simple test problems. The implementation has shown that the processing elements of the artificial neuron are small and simple, but that there can be a significant overhead for the encoding of information into the stochastic pulse rate domain. The stochastic artificial neuron has the capability of on-line weight adaption. The implementation of reinforcement schemes using the stochastic neuron as a basic element are discussed.

The present work was carried out to study the properties of both natural rubber composites reinforced by short polyaramid fibres and natural rubber laminates reinforced by filament polyaramid. Initial work was concentrated to obtain a suitable direct bonding agent and curing system for high modulus nylon 6,6 which was used as the control fibre. Two different types of bonding and curing systems were established as useful for bonding polyaramid to rubbers namely resorcinol-hexamethylene tetraminesilica (HRH) and sulphur, and a blocked diisocyanate in conjunction with a diurethane-crosslinker. Each system was found to offer a different type of adhesion, the HRH formed physical bonds and the NCO/urethane chemical links between fibre and rubber. When the proportions of these bonding agents were optimised in the NR rubber compound vulcanizate properties were significantly improved. Investigation established that for polyaramid fibres the optimum aspect ratio of approximately 1250 was found to give maximum composite reinforcement as judged by the tensile properties of the composite. The experimentally found optimum of 4 parts of blocked diisocyanate was used for extensive investigations of composites properties with respect to fibre orientation and content. Further studies on composites with polyaramid short fibres demonstrated the necessity to optimize the proportions of blocked diisocyanate in the composite to achieve maximum strength properties, e. g. unoptimised with 30% fibre gave a tensile strength of 26.7 MPa whereas optimised, with 30% fibre, gave a tensile strength of 34.4 MPa. Post treatment of composites and laminates by heating or with high energy radiation, using a Co60 source, improved the quality of adhesion still further by the formation of more chemical links between polyaramid and rubber though some degradation of rubber strength simultaneously took place. The bonding agent, diisocyanate, was found unsuitable for use in a sulphur curing system, and was considered to form complexes with the accelerator DCBS and sulphur, to react with zinc salts, and further to act as a crosslinker for rubber. It was therefore found necessary to vulcanize the rubber with a diurethane crosslinker which did not show cure interference with the diisocyanate and the latter then functions primarily as a bonding agent though a minor proportion is shown to simultaneous crosslink the rubber. The Resorcinol-hexamethylene tetramine-silica system. was found to contribute a type of physical adhesion, possibly that of hydrogen bonding, between polyaramid and rubber since this resin formed insitu during vulcanization of rubber was found to crosslink the rubber; reaction mechanisms for both these bonding systems are postulated.

Typescript. / Thesis (Ph. D.)--University of Hawaii at Manoa, 1979. / Bibliography: leaves [70]-75. / Microfiche. / 75 leaves ill. 29 cm