Temporally adaptive estimation of logistic classifiers on data streams

Modern technology has allowed real-time data collection in a variety of domains, ranging from environmental monitoring to healthcare. Consequently, there is a growing need for algorithms capable of performing inferential tasks in an online manner, continuously revising their estimates to reflect the current status of the underlying process. In particular, we are interested in constructing online and temporally adaptive classifiers capable of handling the possibly drifting decision boundaries arising in streaming environments. We first make a quadratic approximation to the log-likelihood that yields a recursive algorithm for fitting logistic regression online. We then suggest a novel way of equipping this framework with self-tuning forgetting factors. The resulting scheme is capable of tracking changes in the underlying probability distribution, adapting the decision boundary appropriately and hence maintaining high classification accuracy in dynamic or unstable environments. We demonstrate the scheme's effectiveness in both real and simulated streaming environments. © Springer-Verlag 2009.

The production of separate streams of pure hydrogen and carbon dioxide from coal via an iron-oxide redox cycle

A chemical looping process using the redox reactions of iron oxide has been used to produce separate streams of pure H2 and CO2 from a solid fuel. An iron oxide carrier prepared using a mechanical mixing technique and comprised of 100wt.% Fe2O3 was used. It was demonstrated that hydrogen can be produced from three representative coals - a Russian bituminous, a German lignite and a UK sub-bituminous coal. Depending on the fuel, pure H2 with [CO] ≲50vol.ppm can be obtained from the proposed process. The cyclic stability of the iron oxide carrier was not adversely affected by contaminants found in syngas which are gaseous above 273K. Stable quantities of H2 were produced over five cycles for all three coals investigated. Independent of the fuel, SO2 was not formed during the oxidation with steam, i.e. the produced H2 was not contaminated with SO2. Since oxidation with air removes contaminants and generates useful heat and pure N2 for purging, it should be included in the operating cycle. Overall, it was demonstrated that the proposed process may be an attractive approach to upgrade crude syngas produced by the gasification of low-rank coals to pure H2, representing a substantial increase in calorific value, whilst simultaneous capturing CO2, a greenhouse gas. © 2010 Elsevier B.V.

Comparing the sustainability of an innovative dry sanitation (Ecosan) system in China to a conventional waterborne sanitation system

The use of sustainability indicators for evaluating sanitation systems is applied to the Erdos Eco- Town Project (EETP) in China for illustration. The EETP is the largest urban settlement in the world employing ecological sanitation, which incorporates separation of waste streams, dry toilets, and resource recovery. The EETP’s dry sanitation system is compared against the Dongsheng District’s conventional sewer and centralised STP. The two systems are compared based on technological, environmental, economic, and societal indicators. Overall, the two systems perform reasonably well from a technological perspective. The conventional system performs significantly better than the dry system with regards to land and energy requirements, and global warming potential; it also performs better based on freshwater aquatic and terrestrial ecotoxicity potentials, but by a smaller margin. The dry system has superior environmental performance based on water consumption, eutrophication potential, and nutrient and organic matter recovery. The dry system is a more costly system as it requires greater infrastructure and higher operational costs, and does not benefit from economies of scale. The waterborne system performs better based on the societal indicators largely because it is a well-established system.