GYAN: A methodology for rule extraction from artificial neural networks

Artificial neural network (ANN) learning methods provide a robust and non-linear approach to approximating the target function for many classification, regression and clustering problems. ANNs have demonstrated good predictive performance in a wide variety of practical problems. However, there are strong arguments as to why ANNs are not sufficient for the general representation of knowledge. The arguments are the poor comprehensibility of the learned ANN, and the inability to represent explanation structures. The overall objective of this thesis is to address these issues by: (1) explanation of the decision process in ANNs in the form of symbolic rules (predicate rules with variables); and (2) provision of explanatory capability by mapping the general conceptual knowledge that is learned by the neural networks into a knowledge base to be used in a rule-based reasoning system. A multi-stage methodology GYAN is developed and evaluated for the task of extracting knowledge from the trained ANNs. The extracted knowledge is represented in the form of restricted first-order logic rules, and subsequently allows user interaction by interfacing with a knowledge based reasoner. The performance of GYAN is demonstrated using a number of real world and artificial data sets. The empirical results demonstrate that: (1) an equivalent symbolic interpretation is derived describing the overall behaviour of the ANN with high accuracy and fidelity, and (2) a concise explanation is given (in terms of rules, facts and predicates activated in a reasoning episode) as to why a particular instance is being classified into a certain category.

Conducting statistical tests of hypotheses : five common misconceptions found in transportation research

Most statistical methods use hypothesis testing. Analysis of variance, regression, discrete choice models, contingency tables, and other analysis methods commonly used in transportation research share hypothesis testing as the means of making inferences about the population of interest. Despite the fact that hypothesis testing has been a cornerstone of empirical research for many years, various aspects of hypothesis tests commonly are incorrectly applied, misinterpreted, and ignored—by novices and expert researchers alike. On initial glance, hypothesis testing appears straightforward: develop the null and alternative hypotheses, compute the test statistic to compare to a standard distribution, estimate the probability of rejecting the null hypothesis, and then make claims about the importance of the finding. This is an oversimplification of the process of hypothesis testing. Hypothesis testing as applied in empirical research is examined here. The reader is assumed to have a basic knowledge of the role of hypothesis testing in various statistical methods. Through the use of an example, the mechanics of hypothesis testing is first reviewed. Then, five precautions surrounding the use and interpretation of hypothesis tests are developed; examples of each are provided to demonstrate how errors are made, and solutions are identified so similar errors can be avoided. Remedies are provided for common errors, and conclusions are drawn on how to use the results of this paper to improve the conduct of empirical research in transportation.

Sodium removal from Maramarua CSG waters using Ngakuru zeolites

Coal seam gas (CSG) waters are a by-product of natural gas extraction from un derground coal seams. The main issue with these waters is their elevated sodium content, which in conjunction with their low calcium and magnesium concentrations can generate soil infiltration problems in the long run , as well as short term toxicity effects in plants due to the sodium ion itself. Zeolites are minerals having a porous structure, crystalline characteristics, and an alumino-silicate configuration resulting in an overall negative charge which is balanced by loosely held cations. In New Zealand, Ngakuru zeolites have been mined for commercial use in wastewater treatment applications, cosmetics, and pet litter. This research focuses on assessing the capacity of Ngakuru zeolites to reduce sodium concentrations of CSG waters from Maramarua. Batch and column test (flow through) experiments revealed that Ngakuru zeolites are capable of sorbing sodium cations from concentrated solutions of sodium. In b atch tests, the sodium adsorption capacity ranged from 5.0 to 34.3meq/100g depending on the solution concentration and on the number of times the zeolite had been regenerated. Regeneration with CaCl2 was foun d to be effective. The calculated sodium adsorption capacity of Ngakuru zeolites under flow-through conditions ranged from 11 to 42meq/100g depending on the strength of the solution being treated and on w hether the zeolites had been previously regenerated. The slow kinetics and low cost of the zeolities, coupled with potentially remote sites for gas extraction, could make semi-batch operational processes without regeneration more favourable than in more industrial ion exchange situations.

Business Process Model Merging : An Approach to Business Process Consolidation

This paper addresses the problem of constructing consolidated business process models out of collections of process models that share common fragments. The paper considers the construction of unions of multiple models (called merged models) as well as intersections (called digests). Merged models are intended for analysts who wish to create a model that subsumes a collection of process models - typically representing variants of the same underlying process - with the aim of replacing the variants with the merged model. Digests, on the other hand, are intended for analysts who wish to identify the most recurring fragments across a collection of process models, so that they can focus their efforts on optimizing these fragments. The paper presents an algorithm for computing merged models and an algorithm for extracting digests from a merged model. The merging and digest extraction algorithms have been implemented and tested against collections of process models taken from multiple application domains. The tests show that the merging algorithm produces compact models and scales up to process models containing hundreds of nodes. Furthermore, a case study conducted in a large insurance company has demonstrated the usefulness of the merging and digest extraction operators in a practical setting.

Prospective, non-randomized phase 2 clinical trial of carboplatin plus paclitaxel with sequential radical pelvic radiotherapy for uterine papillary serous carcinoma

Objective Uterine Papillary Serous Carcinoma (UPSC) is uncommon and accounts for less than 5% of all uterine cancers. Therefore the majority of evidence about the benefits of adjuvant treatment comes from retrospective case series. We conducted a prospective multi-centre non-randomized phase 2 clinical trial using four cycles of adjuvant paclitaxel plus carboplatin chemotherapy followed by pelvic radiotherapy, in order to evaluate the tolerability and safety of this approach. Methods This trial enrolled patients with newly diagnosed, previously untreated patients with stage 1b-4 (FIGO-1988) UPSC with a papillary serous component of at least 30%. Paclitaxel (175 mg/m2) and carboplatin (AUC 6) were administered on day 1 of each 3-week cycle for 4 cycles. Chemotherapy was followed by external beam radiotherapy to the whole pelvis (50.4 Gy over 5.5 weeks). Completion and toxicity of treatment (Common Toxicity Criteria, CTC) and quality of life measures were the primary outcome indicators. Results Twenty-nine of 31 patients completed treatment as planned. Dose reduction was needed in 9 patients (29%), treatment delay in 7 (23%), and treatment cessation in 2 patients (6.5%). Hematologic toxicity, grade 3 or 4 occurred in 19% (6/31) of patients. Patients' self-reported quality of life remained stable throughout treatment. Thirteen of the 29 patients with stages 1–3 disease (44.8%) recurred (average follow up 28.1 months, range 8–60 months). Conclusion This multimodal treatment is feasible, safe and tolerated reasonably well and would be suitable for use in multi-institutional prospective randomized clinical trials incorporating novel therapies in patients with UPSC.

A Survey of Viruses Affecting Dry Bean and Cowpea in Lebanon

A survey was conducted to identify viruses affecting dry bean (Phaseolus vulgaris) and cowpea (Vigna unguiculata) in Lebanon. Three hundred and thirty four samples exhibiting virus-like symptoms were collected from 13 different locations during the fall growing season of 1984. Samples were stored at 20 deg C until they were tested by the enzyme-linked immunosorbent assay (ELISA) for the presence of blackeye cowpea mosaic virus (B1CMV), bean yellow mosaic virus (BYMV), bean common mosaic virus (BCMV) and cucumber mosaic virus (CMV). In preliminary tests, the extraction buffer 0.1M phosphate + O.1MEDTA, pH 7.4 was found to be far better than the standard extraction buffer and, accordingly, was used for virus extraction for all field samples. Results obtained indicated that around 50% of the bean samples tested were infected with B1CMV. Incidence of BCMV, BYMV and CMV in the samples tested were 4,4 and 1.7%, respectively. B1CMV was detected in 10 locations, whereas, BYMV, BCMV and CMV were found in 1,4 and 4 locations, respectively. Mixed infections such as BCMV, BICMV, BCMV+CMV, BYMV+CMV and BICMV+BCMV+CMV were detected. In 35% of the samples assayed, the causal virus was not identified

Simple Serological Assays for Detecting Rice Tungro Viruses

An attempt was made to produce sensitive and specific polyclonal antisera against the viruses causing rice tungro disease, and to assess their potential for use in simple diagnostic tests. Using a multiple, sequential injection procedure, seven batches of polyclonal antisera against rice tungro bacilliform virus (RTBV) and rice tungro spherical virus (RTSV) were produced. These were characterized for their sensitivity and specificity using ring-interface precipitin test and double antibody sandwich (DAS) ELISA. Thirty-one weeks after the first immunization, antiserum batch B6b for RTBV showed the highest ring interface titer (DEP = 1:1920). For RTSV, batches S3, S4b and S5b all had similar titres (DEP = 1:640). In DAS-ELISA, however, significant differences among purified antisera (IgG) batches were observed only at IgG dilution of 10-3. At that dilution, IgGB4b showed the greatest sensitivity, while IgGS3 showed greatest sensitivity for RTSV. When all IgG batches were tested against 11 tungro field isolates (dual RTBV-RTSV infections) at sample dilution of 1:10, IgGB4b and IgGB6b for RTBV and IgGS3 and IgGS6b for RTSV performed equally well. However, after cross adsorption with healthy plant extracts in a specially prepared healthy plant-Sepharose affinity column, only IgGB6b could be used specifically to detect RTBV in a simple tissue-print assay.

Last shall be first : a field study of biases in sequential performance evaluation on the Idol series

When performances are evaluated they are very often presented in a sequential order. Previous research suggests that the sequential presentation of alternatives may induce systematic biases in the way performances are evaluated. Such a phenomenon has been scarcely studied in economics. Using a large dataset of performance evaluation in the Idol series (N=1522), this paper presents new evidence about the systematic biases in sequential evaluation of performances and the psychological phenomena at the origin of these biases.

Sequential development of algebra knowledge : a cognitive analysis

Learning to operate algebraically is a complex process that is dependent upon extending arithmetic knowledge to the more complex concepts of algebra. Current research has shown a gap between arithmetic and algebraic knowledge and suggests a pre-algebraic level as a step between the two knowledge types. This paper examines arithmetic and algebraic knowledge from a cognitive perspective in an effort to determine what constitutes a pre-algebraic level of understanding. Results of a longitudinal study designed to investigate students' readiness for algebra are presented. Thirty-three students in Grades 7, 8, and 9 participated. A model for the transition from arithmetic to pre-algebra to algebra is proposed and students' understanding of relevant knowledge is discussed.