The sample complexity of pattern classification with neural networks: the size of the weights is more important than the size of the network

Sample complexity results from computational learning theory, when applied to neural network learning for pattern classification problems, suggest that for good generalization performance the number of training examples should grow at least linearly with the number of adjustable parameters in the network. Results in this paper show that if a large neural network is used for a pattern classification problem and the learning algorithm finds a network with small weights that has small squared error on the training patterns, then the generalization performance depends on the size of the weights rather than the number of weights. For example, consider a two-layer feedforward network of sigmoid units, in which the sum of the magnitudes of the weights associated with each unit is bounded by A and the input dimension is n. We show that the misclassification probability is no more than a certain error estimate (that is related to squared error on the training set) plus A3 √((log n)/m) (ignoring log A and log m factors), where m is the number of training patterns. This may explain the generalization performance of neural networks, particularly when the number of training examples is considerably smaller than the number of weights. It also supports heuristics (such as weight decay and early stopping) that attempt to keep the weights small during training. The proof techniques appear to be useful for the analysis of other pattern classifiers: when the input domain is a totally bounded metric space, we use the same approach to give upper bounds on misclassification probability for classifiers with decision boundaries that are far from the training examples.

Bounded parameter Markov Decision Processes with average reward criterion

Bounded parameter Markov Decision Processes (BMDPs) address the issue of dealing with uncertainty in the parameters of a Markov Decision Process (MDP). Unlike the case of an MDP, the notion of an optimal policy for a BMDP is not entirely straightforward. We consider two notions of optimality based on optimistic and pessimistic criteria. These have been analyzed for discounted BMDPs. Here we provide results for average reward BMDPs. We establish a fundamental relationship between the discounted and the average reward problems, prove the existence of Blackwell optimal policies and, for both notions of optimality, derive algorithms that converge to the optimal value function.

Overcoming the complexity of diagnostic problems due to sensor network architecture

In fault detection and diagnostics, limitations coming from the sensor network architecture are one of the main challenges in evaluating a system’s health status. Usually the design of the sensor network architecture is not solely based on diagnostic purposes, other factors like controls, financial constraints, and practical limitations are also involved. As a result, it quite common to have one sensor (or one set of sensors) monitoring the behaviour of two or more components. This can significantly extend the complexity of diagnostic problems. In this paper a systematic approach is presented to deal with such complexities. It is shown how the problem can be formulated as a Bayesian network based diagnostic mechanism with latent variables. The developed approach is also applied to the problem of fault diagnosis in HVAC systems, an application area with considerable modeling and measurement constraints.

Narratives of complexity : an ethnographic exploration of resilience and wellbeing among single refugee women in Brisbane

This study aimed to explore resilience and wellbeing among a group of eight refugee women originating from several countries (mainly African) and living in Brisbane, most of whom were single mothers. To challenge mostly quantitative and gender-blind explorations of mental health concepts among refugee groups, the project sought an emic and contextual understanding of resilience and wellbeing. Established perspectives, while useful, tend to overlook the complexities of refugee mental health experiences and can neglect the dense nature of individual stories. The purpose of my study was to contest relatively simplistic narratives of mental health constructs that tend to dominate migrant and refugee studies and influence practice paradigms in the human services field. In this ethnographic exploration of mental health constructs conducted in 2008 and 2009, the use of in-depth interviews, participant observations, and visual ethnographic elements provided an opportunity for refugee women to tell their own stories. The participants’ unique narratives of pre- and post-migration experiences, shaped by specific gender, age, social, cultural and political aspects prevailing in their lives, yielded ‘thick’ ethnographic description (Geertz, 1973) of their social worlds. The findings explored in this study, namely language issues, the impact of community dynamics, and the single status of refugee women, clearly demonstrate that mental health constructs are fluid, multifaceted and complex in reality. In fact, language, community dynamics, and being a single mother, represented both opportunities and barriers in the lives of participants. In some contexts, these factors were conducive to resilience and wellbeing, while in other circumstances, these three elements acted as a hindrance to positive mental health outcomes. There are multiple dimensions to the findings, signifying that the social worlds of refugee women cannot be simplified using set definitions and neat notions of resilience and wellbeing. Instead, the intricacies and complexities embedded in the mundane of the everyday highlight novel conceptualisations of resilience and wellbeing. Based on the particular circumstances of single refugee mothers, whose experiences differ from that of married women, this thesis presents novel articulations of mental health constructs, as an alternative view to existing trends in the literature on refugee issues. Rich and multi-dimensional meanings associated with the socio-cultural determinants of mental health emerged in the process. This thesis’ findings highlight a significant gap in diasporic studies as well as simplistic assumptions about refugee women’s resettlement experiences. Single refugee women’s distinct issues are so complex and dense, that a contextual approach is critical to yield accurate depictions of their circumstances. It is therefore essential to understand refugee lived experiences within broader socio-political contexts to truly appreciate the depth of these narratives. In this manner, critical aspects salient to refugee journeys can inform different understandings of resilience, wellbeing and mental health, and shape contemporary policy and human service practice paradigms.

Arterial travel time estimation : revisiting the classical procedure

Travel time is an important network performance measure and it quantifies congestion in a manner easily understood by all transport users. In urban networks, travel time estimation is challenging due to number of reasons such as, fluctuations in traffic flow due to traffic signals, significant flow to/from mid link sinks/sources, etc. The classical analytical procedure utilizes cumulative plots at upstream and downstream locations for estimating travel time between the two locations. In this paper, we discuss about the issues and challenges with classical analytical procedure such as its vulnerability to non conservation of flow between the two locations. The complexity with respect to exit movement specific travel time is discussed. Recently, we have developed a methodology utilising classical procedure to estimate average travel time and its statistic on urban links (Bhaskar, Chung et al. 2010). Where, detector, signal and probe vehicle data is fused. In this paper we extend the methodology for route travel time estimation and test its performance using simulation. The originality is defining cumulative plots for each exit turning movement utilising historical database which is self updated after each estimation. The performance is also compared with a method solely based on probe (Probe-only). The performance of the proposed methodology has been found insensitive to different route flow, with average accuracy of more than 94% given a probe per estimation interval which is more than 5% increment in accuracy with respect to Probe-only method.

Daily average temperature and mortality among the elderly : a meta-analysis and systematic review of epidemiological evidence

The impact of climate change on the health of vulnerable groups such as the elderly has been of increasing concern. However, to date there has been no meta-analysis of current literature relating to the effects of temperature fluctuations upon mortality amongst the elderly. We synthesised risk estimates of the overall impact of daily mean temperature on elderly mortality across different continents. A comprehensive literature search was conducted using MEDLINE and PubMed to identify papers published up to December 2010. Selection criteria including suitable temperature indicators, endpoints, study-designs and identification of threshold were used. A two-stage Bayesian hierarchical model was performed to summarise the percent increase in mortality with a 1°C temperature increase (or decrease) with 95% confidence intervals in hot (or cold) days, with lagged effects also measured. Fifteen studies met the eligibility criteria and almost 13 million elderly deaths were included in this meta-analysis. In total, there was a 2-5% increase for a 1°C increment during hot temperature intervals, and a 1-2 % increase in all-cause mortality for a 1°C decrease during cold temperature intervals. Lags of up to 9 days in exposure to cold temperature intervals were substantially associated with all-cause mortality, but no substantial lagged effects were observed for hot intervals. Thus, both hot and cold temperatures substantially increased mortality among the elderly, but the magnitude of heat-related effects seemed to be larger than that of cold effects within a global context.

Quantification of particle emission characteristics and development of an emission model for use in transport microenvironments affected by traffic emissions

Vehicle emitted particles are of significant concern based on their potential to influence local air quality and human health. Transport microenvironments usually contain higher vehicle emission concentrations compared to other environments, and people spend a substantial amount of time in these microenvironments when commuting. Currently there is limited scientific knowledge on particle concentration, passenger exposure and the distribution of vehicle emissions in transport microenvironments, partially due to the fact that the instrumentation required to conduct such measurements is not available in many research centres. Information on passenger waiting time and location in such microenvironments has also not been investigated, which makes it difficult to evaluate a passenger’s spatial-temporal exposure to vehicle emissions. Furthermore, current emission models are incapable of rapidly predicting emission distribution, given the complexity of variations in emission rates that result from changes in driving conditions, as well as the time spent in driving condition within the transport microenvironment. In order to address these scientific gaps in knowledge, this work conducted, for the first time, a comprehensive statistical analysis of experimental data, along with multi-parameter assessment, exposure evaluation and comparison, and emission model development and application, in relation to traffic interrupted transport microenvironments. The work aimed to quantify and characterise particle emissions and human exposure in the transport microenvironments, with bus stations and a pedestrian crossing identified as suitable research locations representing a typical transport microenvironment. Firstly, two bus stations in Brisbane, Australia, with different designs, were selected to conduct measurements of particle number size distributions, particle number and PM2.5 concentrations during two different seasons. Simultaneous traffic and meteorological parameters were also monitored, aiming to quantify particle characteristics and investigate the impact of bus flow rate, station design and meteorological conditions on particle characteristics at stations. The results showed higher concentrations of PN20-30 at the station situated in an open area (open station), which is likely to be attributed to the lower average daily temperature compared to the station with a canyon structure (canyon station). During precipitation events, it was found that particle number concentration in the size range 25-250 nm decreased greatly, and that the average daily reduction in PM2.5 concentration on rainy days compared to fine days was 44.2 % and 22.6 % at the open and canyon station, respectively. The effect of ambient wind speeds on particle number concentrations was also examined, and no relationship was found between particle number concentration and wind speed for the entire measurement period. In addition, 33 pairs of average half-hourly PN7-3000 concentrations were calculated and identified at the two stations, during the same time of a day, and with the same ambient wind speeds and precipitation conditions. The results of a paired t-test showed that the average half-hourly PN7-3000 concentrations at the two stations were not significantly different at the 5% confidence level (t = 0.06, p = 0.96), which indicates that the different station designs were not a crucial factor for influencing PN7-3000 concentrations. A further assessment of passenger exposure to bus emissions on a platform was evaluated at another bus station in Brisbane, Australia. The sampling was conducted over seven weekdays to investigate spatial-temporal variations in size-fractionated particle number and PM2.5 concentrations, as well as human exposure on the platform. For the whole day, the average PN13-800 concentration was 1.3 x 104 and 1.0 x 104 particle/cm3 at the centre and end of the platform, respectively, of which PN50-100 accounted for the largest proportion to the total count. Furthermore, the contribution of exposure at the bus station to the overall daily exposure was assessed using two assumed scenarios of a school student and an office worker. It was found that, although the daily time fraction (the percentage of time spend at a location in a whole day) at the station was only 0.8 %, the daily exposure fractions (the percentage of exposures at a location accounting for the daily exposure) at the station were 2.7% and 2.8 % for exposure to PN13-800 and 2.7% and 3.5% for exposure to PM2.5 for the school student and the office worker, respectively. A new parameter, “exposure intensity” (the ratio of daily exposure fraction and the daily time fraction) was also defined and calculated at the station, with values of 3.3 and 3.4 for exposure to PN13-880, and 3.3 and 4.2 for exposure to PM2.5, for the school student and the office worker, respectively. In order to quantify the enhanced emissions at critical locations and define the emission distribution in further dispersion models for traffic interrupted transport microenvironments, a composite line source emission (CLSE) model was developed to specifically quantify exposure levels and describe the spatial variability of vehicle emissions in traffic interrupted microenvironments. This model took into account the complexity of vehicle movements in the queue, as well as different emission rates relevant to various driving conditions (cruise, decelerate, idle and accelerate), and it utilised multi-representative segments to capture the accurate emission distribution for real vehicle flow. This model does not only helped to quantify the enhanced emissions at critical locations, but it also helped to define the emission source distribution of the disrupted steady flow for further dispersion modelling. The model then was applied to estimate particle number emissions at a bidirectional bus station used by diesel and compressed natural gas fuelled buses. It was found that the acceleration distance was of critical importance when estimating particle number emission, since the highest emissions occurred in sections where most of the buses were accelerating and no significant increases were observed at locations where they idled. It was also shown that emissions at the front end of the platform were 43 times greater than at the rear of the platform. The CLSE model was also applied at a signalled pedestrian crossing, in order to assess increased particle number emissions from motor vehicles when forced to stop and accelerate from rest. The CLSE model was used to calculate the total emissions produced by a specific number and mix of light petrol cars and diesel passenger buses including 1 car travelling in 1 direction (/1 direction), 14 cars / 1 direction, 1 bus / 1 direction, 28 cars / 2 directions, 24 cars and 2 buses / 2 directions, and 20 cars and 4 buses / 2 directions. It was found that the total emissions produced during stopping on a red signal were significantly higher than when the traffic moved at a steady speed. Overall, total emissions due to the interruption of the traffic increased by a factor of 13, 11, 45, 11, 41, and 43 for the above 6 cases, respectively. In summary, this PhD thesis presents the results of a comprehensive study on particle number and mass concentration, together with particle size distribution, in a bus station transport microenvironment, influenced by bus flow rates, meteorological conditions and station design. Passenger spatial-temporal exposure to bus emitted particles was also assessed according to waiting time and location along the platform, as well as the contribution of exposure at the bus station to overall daily exposure. Due to the complexity of the interrupted traffic flow within the transport microenvironments, a unique CLSE model was also developed, which is capable of quantifying emission levels at critical locations within the transport microenvironment, for the purpose of evaluating passenger exposure and conducting simulations of vehicle emission dispersion. The application of the CLSE model at a pedestrian crossing also proved its applicability and simplicity for use in a real-world transport microenvironment.

Complexity, evolution, and the structure of demand

This chapter argues that evolutionary economics should be founded upon complex systems theory rather than neo-Darwinian analogies concerning natural selection, which focus on supply side considerations and competition amongst firms and technologies. It suggests that conceptions such as production and consumption functions should be replaced by network representations, in which the preferences or, more correctly, the aspirations of consumers are fundamental and, as such, the primary drivers of economic growth. Technological innovation is viewed as a process that is intermediate between these aspirational networks, and the organizational networks in which goods and services are produced. Consumer knowledge becomes at least as important as producer knowledge in determining how economic value is generated. It becomes clear that the stability afforded by connective systems of rules is essential for economic flexibility to exist, but that too many rules result in inert and structurally unstable states. In contrast, too few rules result in a more stable state, but at a low level of ordered complexity. Economic evolution from this perspective is explored using random and scale free network representations of complex systems.

Combined time- and space-resolved Raman spectrometer for the non-invasive depth profiling of chemical hazards

A time-resolved inverse spatially offset Raman spectrometer was constructed for depth profiling of Raman-active substances under both the lab and the field environments. The system operating principles and performance are discussed along with its advantages relative to traditional continuous wave spatially offset Raman spectrometer. The developed spectrometer uses a combination of space- and time-resolved detection in order to obtain high-quality Raman spectra from substances hidden behind coloured opaque surface layers, such as plastic and garments, with a single measurement. The time-gated spatially offset Raman spectrometer was successfully used to detect concealed explosives and drug precursors under incandescent and fluorescent background light as well as under daylight. The average screening time was 50 s per measurement. The excitation energy requirements were relatively low (20 mW) which makes the probe safe for screening hazardous substances. The unit has been designed with nanosecond laser excitation and gated detection, making it of lower cost and complexity than previous picosecond-based systems, to provide a functional platform for in-line or in-field sensing of chemical substances.

Measuring complexity for building projects : a Delphi study

Purpose---The aim of this study is to identify complexity measures for building projects in the People’s Republic of China (PRC). Design/Methodology/Approach---A three-round of Delphi questionnaire survey was conducted to identify the key parameters that measure the degree of project complexity. A complexity index (CI) was developed based on the identified measures and their relative importance. Findings---Six key measures of project complexity have been identified, which include, namely (1) building structure & function; (2) construction method; (3) the urgency of the project schedule; (4) project size/scale; (5) geological condition; and (6) neighboring environment. Practical implications---These complexity measures help stakeholders assess degrees of project complexity and better manage the potential risks that might be induced to different levels of project complexity. Originality/Value---The findings provide insightful perspectives to define and understand project complexity. For stakeholders, understanding and addressing the complexity help to improve project planning and implementation.

Risk, uncertainty and complexity in science education

Increasingly societies and their governments are facing important social issues that have science and technology as key features. A number of these socio-scientific issues have two features that distinguish them from the restricted contexts in which school science has traditionally been presented. Some of their science is uncertain and scientific knowledge is not the only knowledge involved. As a result, the concepts of uncertainty, risk and complexity become essential aspects of the science underlying these issues. In this chapter we discuss the nature and role of these concepts in the public understanding of science and consider their links with school science. We argue that these same concepts and their role in contemporary scientific knowledge need to be addressed in school science curricula. The new features for content, pedagogy and assessment of this urgent challenge for science educators are outlined. These will be essential if the goal of science education for citizenship is to be achieved with our students, who will increasingly be required to make personal and collective decisions on issues involving science and technology.