University of New Brunswick

High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for fine-tuning the weights in such "autoencoder" networks, but this works well only if the initial weights are close to a good solution. We describe an effective way of initializing the weights that allows deep autoencoder networks to learn low-dimensional codes that work much better than principal components analysis as a tool to reduce the dimensionality of data.

During the last decade, anomaly detection has attracted the attention of many researchers to overcome the weakness of signature-based IDSs in detecting novel attacks, and KDDCUP'99 is the mostly widely used data set for the evaluation of these systems. Having conducted a statistical analysis on this data set, we found two important issues which highly affects the performance of evaluated systems, and results in a very poor evaluation of anomaly detection approaches. To solve these issues, we have proposed a new data set, NSL-KDD, which consists of selected records of the complete KDD data set and does not suffer from any of mentioned shortcomings.

We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder, and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations of molecules allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in a set of molecules with fewer that nine heavy atoms.

PURPOSE: Emerging evidence suggests muscle depletion predicts survival of patients with cancer. PATIENTS AND METHODS: At a cancer center in Alberta, Canada, consecutive patients with cancer (lung or GI; N = 1,473) were assessed at presentation for weight loss history, lumbar skeletal muscle index, and mean muscle attenuation (Hounsfield units) by computed tomography (CT). Univariate and multivariate analyses were conducted. Concordance (c) statistics were used to test predictive accuracy of survival models. RESULTS: Body mass index (BMI) distribution was 17% obese, 35% overweight, 36% normal weight, and 12% underweight. Patients in all BMI categories varied widely in weight loss, muscle index, and muscle attenuation. Thresholds defining associations between these three variables and survival were determined using optimal stratification. High weight loss, low muscle index, and low muscle attenuation were independently prognostic of survival. A survival model containing conventional covariates (cancer diagnosis, stage, age, performance status) gave a c statistic of 0.73 (95% CI, 0.67 to 0.79), whereas a model ignoring conventional variables and including only BMI, weight loss, muscle index, and muscle attenuation gave a c statistic of 0.92 (95% CI, 0.88 to 0.95; P < .001). Patients who possessed all three of these poor prognostic variables survived 8.4 months (95% CI, 6.5 to 10.3), regardless of whether they presented as obese, overweight, normal weight, or underweight, in contrast to patients who had none of these features, who survived 28.4 months (95% CI, 24.2 to 32.6; P < .001). CONCLUSION: CT images reveal otherwise occult muscle depletion. Patients with cancer who are cachexic by the conventional criterion (involuntary weight loss) and by two additional criteria (muscle depletion and low muscle attenuation) share a poor prognosis, regardless of overall body weight.

ABSTRACT Estimation of latent ability using the entire response pattern of free‐response items is discussed, first in the general case and then in the case where the items are scored in a graded way, especially when the thinking process required for solving each item is assumed to be homogeneous. The maximum likelihood estimator, the Bayes modal estimator, and the Bayes estimator obtained by using the mean‐square error multiplied by the density function of the latent variate as the loss function are taken as our estimators. Sufficient conditions for the existence of a unique maximum likelihood estimator and a unique Bayes modal estimator are formulated with respect to an individual item rather than with respect to a whole set of items, which are useful especially in the situation where we are free to choose optimal items for a particular examinee out of the item library in which a sufficient number of items are stored with reliable quality controls. Advantages of the present methods are investigated by comparing them with those which make use of conventional dichotomous items or test scores, theoretically as well as empirically, in terms of the amounts of information, the standard errors of estimators, and the mean‐square errors of estimators. The utility of the Bayes modal estimator as a computational compromise for the Bayes estimator is also discussed and observed. The relationship between the formula for the item characteristic function and the philosophy of scoring is observed with respect to dichotomous items.

Exposure to ambient fine particulate matter (PM 2.5 ) is a major global health concern. Quantitative estimates of attributable mortality are based on disease-specific hazard ratio models that incorporate risk information from multiple PM 2.5 sources (outdoor and indoor air pollution from use of solid fuels and secondhand and active smoking), requiring assumptions about equivalent exposure and toxicity. We relax these contentious assumptions by constructing a PM 2.5 -mortality hazard ratio function based only on cohort studies of outdoor air pollution that covers the global exposure range. We modeled the shape of the association between PM 2.5 and nonaccidental mortality using data from 41 cohorts from 16 countries—the Global Exposure Mortality Model (GEMM). We then constructed GEMMs for five specific causes of death examined by the global burden of disease (GBD). The GEMM predicts 8.9 million [95% confidence interval (CI): 7.5–10.3] deaths in 2015, a figure 30% larger than that predicted by the sum of deaths among the five specific causes (6.9; 95% CI: 4.9–8.5) and 120% larger than the risk function used in the GBD (4.0; 95% CI: 3.3–4.8). Differences between the GEMM and GBD risk functions are larger for a 20% reduction in concentrations, with the GEMM predicting 220% higher excess deaths. These results suggest that PM 2.5 exposure may be related to additional causes of death than the five considered by the GBD and that incorporation of risk information from other, nonoutdoor, particle sources leads to underestimation of disease burden, especially at higher concentrations.

This monograph is a part of a more comprehensive treatment of estimation of latent traits, when the entire response pattern is used. The fundamental structure of the whole theory comes from the latent trait model, which was initiated by Lazarsfeld as the latent structure analysis [Lazarsfeld, 1959], and also by Lord and others as a theory of mental test scores [Lord, 1952]. Similarities and differences in their mathematical structures and tendencies were discussed by Lazarsfeld [Lazarsfeld, 1960] and the recent book by Lord and Novick with contributions by Birnbaum [Lord &amp; Novick, 1968] provides the dichotomous case of the latent trait model in the context of mental measurement.

urban populations. We begin by defining the state of the art, explaining the science of smart cities. We define six scenarios based on new cities badging themselves as smart, older cities regenerating themselves as smart, the development of science parks, tech cities, and technopoles focused on high technologies, the development of urban services using contemporary ICT, the use of ICT to develop new urban intelligence functions, and the development of online and mobile forms of participation. Seven project areas are then proposed: Integrated Databases for the Smart City, Sensing, Networking and the Impact of New Social Media, Modelling Network Performance, Mobility and Travel Behaviour, Modelling Urban Land Use, Transport and Economic Interactions, Modelling Urban Transactional Activities in Labour and Housing Markets, Decision Support as Urban Intelligence, Participatory Governance and Planning Structures for the Smart City. Finally we anticipate the paradigm shifts that will occur in this research and define a series of key demonstrators which we believe are important to progressing a science of smart cities.

This paper describes a novel approach to the control of a multifunction prosthesis based on the classification of myoelectric patterns. It is shown that the myoelectric signal exhibits a deterministic structure during the initial phase of a muscle contraction. Features are extracted from several time segments of the myoelectric signal to preserve pattern structure. These features are then classified using an artificial neural network. The control signals are derived from natural contraction patterns which can be produced reliably with little subject training. The new control scheme increases the number of functions which can be controlled by a single channel of myoelectric signal but does so in a way which does not increase the effort required by the amputee. Results are presented to support this approach.

We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder, and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations of molecules allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in a set of molecules with fewer that nine heavy atoms.

Meta‐analytic techniques were used to determine which predictor domains and actuarial assessment instruments were the best predictors of adult offender recidivism. One hundred and thirty‐one studies produced 1,141 correlations with recidivism. The strongest predictor domains were criminogenic needs, criminal history/history of antisocial behavior, social achievement, age/gender/race, and family factors. Less robust predictors included intellectual functioning, personal distress factors, and socioeconomic status in the family of origin. Dynamic predictor domains performed at least as well as the static domains. The LSI‐R was identified as the most useful actuarial measure. Recommendations for developing sound assessment practices in corrections are provided.

Better modelling and analysis of problem domains, as opposed to the behaviour of s oflwa re. Development of richer models for capturing and analysing non-functional requirements. Bridging the gap between requirements elicitation approaches based on contextual enquiry and more formal specification and analysis techniques. Better understanding of the impact of software architectural choices on the prioritisation and evolution of requirements. Reuse of requirements models to facilitate the development of system families and the selection of COTS. Multi-disciplinary training for requirements practitioners.

This paper represents an ongoing investigation of dexterous and natural control of upper extremity prostheses using the myoelectric signal (MES). The scheme described within uses pattern recognition to process four channels of MES, with the task of discriminating multiple classes of limb movement. The method does not require segmentation of the MES data, allowing a continuous stream of class decisions to be delivered to a prosthetic device. It is shown in this paper that, by exploiting the processing power inherent in current computing systems, substantial gains in classifier accuracy and response time are possible. Other important characteristics for prosthetic control systems are met as well. Due to the fact that the classifier learns the muscle activation patterns for each desired class for each individual, a natural control actuation results. The continuous decision stream allows complex sequences of manipulation involving multiple joints to be performed without interruption. Finally, minimal storage capacity is required, which is an important factor in embedded control systems.

Autonomous underwater vehicle (AUV) navigation and localization in underwater environments is particularly challenging due to the rapid attenuation of Global Positioning System (GPS) and radio-frequency signals. Underwater communications are low bandwidth and unreliable, and there is no access to a global positioning system. Past approaches to solve the AUV localization problem have employed expensive inertial sensors, used installed beacons in the region of interest, or required periodic surfacing of the AUV. While these methods are useful, their performance is fundamentally limited. Advances in underwater communications and the application of simultaneous localization and mapping (SLAM) technology to the underwater realm have yielded new possibilities in the field. This paper presents a review of the state of the art of AUV navigation and localization, as well as a description of some of the more commonly used methods. In addition, we highlight areas of future research potential.

Graceful choreography for CO 2 and H 2 O One challenge for efficient electrochemical reduction of carbon dioxide (CO 2 ) is that the gas is hydrophobic, but many of its desirable reactions require water (H 2 O). García de Arquer et al. addressed this problem by combining a copper electrocatalyst with an ionomer assembly that intersperses sulfonate-lined paths for the H 2 O with fluorocarbon channels for the CO 2 . The electrode architecture enables production of two-carbon products such as ethylene and ethanol at current densities just over an ampere per square centimeter. Science , this issue p. 661

Naive Bayes is one of the most efficient and effective inductive learning algorithms for machine learning and data mining. Its competitive performance in classifica-tion is surprising, because the conditional independence assumption on which it is based, is rarely true in real-world applications. An open question is: what is the true reason for the surprisingly good performance of naive Bayes in classification? In this paper, we propose a novel explanation on the superb classification performance of naive Bayes. We show that, essentially, the dependence distribution; i.e., how the local dependence of a node distributes in each class, evenly or unevenly, and how the local dependen-cies of all nodes work together, consistently (support-

This article investigates the processes of word of mouth (WOM) within a services purchase decision context. The authors argue that to understand these processes, researchers must examine the role of interpersonal influences in the traditional WOM models based within the noninterpersonal paradigm. As a result of the current investigation, three distinct relations emerge: first, the effect of the noninterpersonal forces (receiver’s expertise, receiver’s perceived risk, and sender’s expertise) on the influence of WOM on service purchase decisions; second, the effect of the interpersonal forces (ties strength and how actively WOM is sought) on the influence of WOM on service purchase decisions; and third, the effects of noninterpersonal forces on interpersonal forces. Managerial implications and avenues for future research are addressed.

A three-stage schema-based information processing model of anxiety is described that involves: (a) the initial registration of a threat stimulus; (b) the activation of a primal threat mode; and (c) the secondary activation of more elaborative and reflective modes of thinking. The defining elements of automatic and strategic processing are discussed with the cognitive bias in anxiety reconceptualized in terms of a mixture of automatic and strategic processing characteristics depending on which stage of the information processing model is under consideration. The goal in the treatment of anxiety is to deactivate the more automatic primal threat mode and to strengthen more constructive reflective modes of thinking. Arguments are presented for the inclusion of verbal mediation as a necessary but not sufficient component in the cognitive and behavioral treatment of anxiety.

A female advantage in school marks is a common finding in education research, and it extends to most course subjects (e.g., language, math, science), unlike what is found on achievement tests. However, questions remain concerning the quantification of these gender differences and the identification of relevant moderator variables. The present meta-analysis answered these questions by examining studies that included an evaluation of gender differences in teacher-assigned school marks in elementary, junior/middle, or high school or at the university level (both undergraduate and graduate). The final analysis was based on 502 effect sizes drawn from 369 samples. A multilevel approach to meta-analysis was used to handle the presence of nonindependent effect sizes in the overall sample. This method was complemented with an examination of results in separate subject matters with a mixed-effects meta-analytic model. A small but significant female advantage (mean d = 0.225, 95% CI [0.201, 0.249]) was demonstrated for the overall sample of effect sizes. Noteworthy findings were that the female advantage was largest for language courses (mean d = 0.374, 95% CI [0.316, 0.432]) and smallest for math courses (mean d = 0.069, 95% CI [0.014, 0.124]). Source of marks, nationality, racial composition of samples, and gender composition of samples were significant moderators of effect sizes. Finally, results showed that the magnitude of the female advantage was not affected by year of publication, thereby contradicting claims of a recent "boy crisis" in school achievement. The present meta-analysis demonstrated the presence of a stable female advantage in school marks while also identifying critical moderators. Implications for future educational and psychological research are discussed.

Abstract The electrochemical reduction of CO 2 is a promising route to convert intermittent renewable energy to storable fuels and valuable chemical feedstocks. To scale this technology for industrial implementation, a deepened understanding of how the CO 2 reduction reaction (CO 2 RR) proceeds will help converge on optimal operating parameters. Here, a techno‐economic analysis is presented with the goal of identifying maximally profitable products and the performance targets that must be met to ensure economic viability—metrics that include current density, Faradaic efficiency, energy efficiency, and stability. The latest computational understanding of the CO 2 RR is discussed along with how this can contribute to the rational design of efficient, selective, and stable electrocatalysts. Catalyst materials are classified according to their selectivity for products of interest and their potential to achieve performance targets is assessed. The recent progress and opportunities in system design for CO 2 electroreduction are described. To conclude, the remaining technological challenges are highlighted, suggesting full‐cell energy efficiency as a guiding performance metric for industrial impact.