Enhancements to the dynamic self-organising neural network

Class Directed Unsupervised Learning (CDUL) is a dynamic self-organising network which has been shown to overcome many of the problems associated with unsupervised learning, thereby yielding performance characteristics superior to similar networks such as counter-propagation and LVQ. In this paper, the CDUL algorithm is developed further, to a point where the original two-phase learning process is combined into a single system of dynamic parameter variation; a training cycle that can then be terminated automatically at a point of zero error over the training set. The ability to improve training times using a FastCDUL algorithm is also explored. The new algorithm, CDUL2, is subsequently applied to the benchmark problem of mine detection given sonar data, and shown to outperform both backpropagation and LVQ in terms of training speed and recall performance. Finally, a measure of computational cost is estimated for both CDUL2 and LVQ training periods, reinforcing the suggested efficiency of CDUL2 over its counterparts.     

Survey of impaired drivers, fatally injured or surviving, who caused fatal highway accidents in Alberta in 1970-72

In 1970-72, 456 (53.4%) of 854 drivers responsible for a motor vehicle accident in which either they or other persons were killed had been drinking beforehand. Of surviving culpable drivers tested for blood alcohol concentration 81.7% were found to be legally impaired; however, only 24.8% of these drivers were tested. Because surviving culpable drivers pose a possible future hazard to highway safety it is important that data on this group be collected and analysed.     

Fish, fishing, and pollutant reduction in the Baltic Sea

The Baltic Sea is heavily polluted yet supports major commercial fisheries for cod (Gadus morhua), herring (Clupea harengus), and sprat (Sprattus sprattus). Emissions of persistent organic pollutants, such as polychlorinated biphenyls (PCBs) and DDT, were high during the 1960s and 1970s, and concentrations in fish and other fauna are still significant. Several models of the fluxes of these pollutants among the water, sediment, and atmosphere have been developed, but these generally omit the roles of fish and fisheries. We show that the standing stock of the most abundant fish species in the Baltic Sea was a sink for 260 kg of PCBs in the late 1980s to early 1990s and that the fishery removed as much or more PCB (31 kg yr(-1))than other budget components (e.g., degradation in the water column). Accounting for fish and fisheries could increase our understanding of the fluxes of pollutants, and banning the discard of highly contaminated organs such as cod liver could be part of the pollution management.     

Mechanochemical synthesis and sintering behaviour of magnesium aluminate spinel

X-ray amorphous precursor phases for the synthesis of spinel (MgAl₂O₄) have been prepared by grinding mixtures of gibbsite (Al(OH)₃) with brucite (Mg(OH)₂) or hydromagnesite (4MgCO₃·Mg(OH)₂·4H₂O). The mechanochemical treatment does not remove any water or carbonate, but converts some of the gibbsite octahedral Al sites into tetrahedral sites and other sites with a ²⁷Al MAS NMR resonance at about 38 ppm. The brucite-derived precursor forms spinel on heating at 850°C, by contrast with unground mixtures which show little spinel formation even at 1250°C. The hydromagnesite-derived precursor transforms at about 850°C into a mixture of spinel and hydrotalcite (Mg₆Al₂(OH)₁₆CO₃·4H₂O), the latter decomposing to spinel and MgO by 1050°C. Spinel derived from the hydromagnesite-containing precursor shows superior pressureless sintering properties at 1400–1600°C, producing a body of 97% theoretical bulk density at 1600°C. Under the same conditions, the brucite-derived spinel sintered to 72% theoretical density and showed a morphology consisting of widely disparate grain sizes.     

Oxygen evolution during the formation and sintering of nickel-manganese oxide spinels for thermistor applications

Nickel-manganese spinel, prepared from 20 wt % NiO and 80 wt % Mn₂O₃, forms at 950°C by the intermediate formation of Mn₃O₄ with evolution of oxygen, determined by mass-spectrometric evolved gas analysis. On heating to higher temperatures, further oxygen is evolved resulting in pore formation and bloating of pressed sintered samples and anomalies in their densification and electrical properties. Thermodynamic considerations and X-ray diffraction intensity measurements suggest that all the Mn²⁺ is located in the tetrahedral spinel sites, this cation configuration remaining unchanged by higher-temperature treatments such as sintering.     

Intra-storm variability in microbial partitioning and microbial loading rates

Association with particles in the water column can have a significant impact on microbial fate and transport. This study analyzed multiple stormwater samples taken throughout the duration of three separate storms (at two different sites) to evaluate the fraction of microbes partitioning to denser "settleable" particles and to examine how partitioning behavior varied over the course of a storm. Intra-storm sampling also allowed for estimates of microbial loading rates (both total and particle-associated) and cumulative storm-induced microbial load. Five different indicator organisms were examined, with the fraction of microbes associated with settleable particles assessed via a calibrated centrifugation method. Partitioning behavior varied across microorganism type, with an average of 40% of fecal coliforms, Escherichia coli, and enterococci associating with settleable particles, compared to approximately 65% of Clostridium perfringens spores and only 13% of total coliphage. Partitioning remained fairly constant for each type of organism throughout storm events. Nonetheless, higher concentrations of both settleable particles and microbes entering the water column soon after the onset of a storm led to higher loading rates of settleable microbes in the storm's earliest stages, a trend that could have important implications for the design of stormwater management structures (e.g., detention basins). Estimates of cumulative storm-induced microbial loading suggested that one day's worth of storm loading can be the equivalent of months, or even years, of dry-weather loading.     

A computer algebra based finite element development environment

A finite element development environment based on the technical computing program Mathematica is described. The environment is used to automatically program standard element formulations and develop new elements with novel features. Source code can also be exported in a format compatible with commercial finite element program user-element facilities. The development environment is demonstrated for three mixed Petrov–Galerkin plane stress elements: a standard formulation, an advanced formulation incorporating rotational degrees of freedom and a standard formulation in which the stiffness matrix is integrated analytically, before being exported as ANSYS user elements. The results presented illustrate the accuracy of the standard mixed formulation element and the enhancement of performance when rotational degrees of freedom are added. Further, the analytically integrated element shows that computational requirements can be greatly reduced when analytical integration schemes are used in the formation.     

Activity Awareness as an Enabler for Communication and Network Building in Construction Design Teams

The paper reports on ongoing research into the facilitation of communication in distributed design teams of the type typically encountered in the construction sector. Ideas around social aspects in such design teams are discussed followed by our previous research efforts as an evolutionary process that has led us to develop in this direction. It is argued that facilitating timely and contextually grounded communication in a distributed environment can help to build the social networks and associated social capital common in collocated design teams but often absent in distributed ones. We describe our framework for providing such facilitation through monitoring and awareness of the information “space” of a construction project. We believe that by leveraging the relationships in the information space, we can go some way to identifying those individuals who would benefit from developing interpersonal relations in their work and further by exploiting those relationships in real time we can accelerate the formation of personal social networks and social capital between those individuals.