Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and deterministic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel metaheuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS metaheuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and determinis- tic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel meta–heuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS meta–heuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

The effects of teaching a mathematics problem solving strategy on the problem solving ability of grade nine students

Forty grade 9 students were selected from a small rural board in southern Ontario. The students were in two classes and were treated as two groups. The treatment group received instruction in the Logical Numerical Problem Solving Strategy every day for 37 minutes over a 6 week period. The control group received instruction in problem solving without this strategy over the same time period. Then the control group received the treat~ent and the treatment group received the instruction without the strategy. Quite a large variance was found in the problem solving ability of students in grade 9. It was also found that the growth of the problem solving ability achievement of students could be measured using growth strands based upon the results of the pilot study. The analysis of the results of the study using t-tests and a MANOVA demonstrated that the teaching of the strategy did not significaritly (at p s 0.05) increase the problem solving achievement of the students. However, there was an encouraging trend seen in the data.

A stochastic dynamic programming approach for pricing options on stock-index futures

The aim of this thesis is to price options on equity index futures with an application to standard options on S&P 500 futures traded on the Chicago Mercantile Exchange. Our methodology is based on stochastic dynamic programming, which can accommodate European as well as American options. The model accommodates dividends from the underlying asset. It also captures the optimal exercise strategy and the fair value of the option. This approach is an alternative to available numerical pricing methods such as binomial trees, finite differences, and ad-hoc numerical approximation techniques. Our numerical and empirical investigations demonstrate convergence, robustness, and efficiency. We use this methodology to value exchange-listed options. The European option premiums thus obtained are compared to Black's closed-form formula. They are accurate to four digits. The American option premiums also have a similar level of accuracy compared to premiums obtained using finite differences and binomial trees with a large number of time steps. The proposed model accounts for deterministic, seasonally varying dividend yield. In pricing futures options, we discover that what matters is the sum of the dividend yields over the life of the futures contract and not their distribution.

Evolutionary Origin and Maintenance of Sociality in the Small Carpenter Bees

Many arthropods exhibit behaviours precursory to social life, including adult longevity, parental care, nest loyalty and mutual tolerance, yet there are few examples of social behaviour in this phylum. The small carpenter bees, genus Ceratina, provide important insights into the early stages of sociality. I described the biology and social behaviour of five facultatively social species which exhibit all of the preadaptations for successful group living, yet present ecological and behavioural characteristics that seemingly disfavour frequent colony formation. These species are socially polymorphic with both / solitary and social nests collected in sympatry. Social colonies consist of two adult females, one contributing both foraging and reproductive effort and the second which remains at the nest as a passive guard. Cooperative nesting provides no overt reproductive benefits over solitary nesting, although brood survival tends to be greater in social colonies. Three main theories explain cooperation among conspecifics: mutual benefit, kin selection and manipulation. Lifetime reproductive success calculations revealed that mutual benefit does not explain social behaviour in this group as social colonies have lower per capita life time reproductive success than solitary nests. Genetic pedigrees constructed from allozyme data indicate that kin selection might contribute to the maintenance of social nesting -, as social colonies consist of full sisters and thus some indirect fitness benefits are inherently bestowed on subordinate females as a result of remaining to help their dominant sister. These data suggest that the origin of sociality in ceratinines has principal costs and the great ecological success of highly eusociallineages occurred well after social origins. Ecological constraints such as resource limitation, unfavourable weather conditions and parasite pressure have long been considered some of the most important selective pressures for the evolution of sociality. I assessed the fitness consequences of these three ecological factors for reproductive success of solitary and social colonies and found that nest sites were not limiting, and the frequency of social nesting was consistent across brood rearing seasons. Local weather varied between seasons but was not correlated with reproductive success. Severe parasitism resulted in low reproductive success and total nest failure in solitary nests. Social colonies had higher reproductive success and were never extirpated by parasites. I suggest that social nesting represents a form of bet-hedging. The high frequency of solitary nests suggests that this is the optimal strategy when parasite pressure is low. However, social colonies have a selective advantage over solitary nesting females during periods of extreme parasite pressure. Finally, the small carpenter bees are recorded from all continents except Antarctica. I constructed the first molecular phylogeny of ceratinine bees based on four gene regions of selected species covering representatives from all continents and ecological regions. Maximum parsimony and Bayesian Inference tree topology and fossil dating support an African origin followed by an Old World invasion and New World radiation. All known Old World ceratinines form social colonies while New World species are largely solitary; thus geography and phylogenetic inertia are likely predictors of social evolution in this genus. This integrative approach not only describes the behaviour of several previously unknown or little-known Ceratina species, bu~ highlights the fact that this is an important, though previously unrecognized, model for studying evolutionary transitions from solitary to social behaviour.

In Search of Convergence: A Co-citation Analysis of Three Sport Management Journals

Responding to a series of articles in sport management literature calling for more diversity in terms of areas of interest or methods, this study warns against the danger of excessively fragmenting this field of research. The works of Kuhn (1962) and Pfeffer (1993) are taken as the basis of an argument that connects convergence with scientific strength. However, being aware of the large number of counterarguments directed at this line of reasoning, a new model of convergence, which focuses on clusters of research contributions with similar areas of interest, methods, and concepts, is proposed. The existence of these clusters is determined with the help of a bibliometric analysis of publications in three sport management journals. This examination determines that there are justified reasons to be concerned about the level of convergence in the field, pointing out to a reduced ability to create large clusters of contributions in similar areas of interest.

Heuristics for the Critical Node Detection Problem in Large Complex Networks

Complex networks have recently attracted a significant amount of research attention due to their ability to model real world phenomena. One important problem often encountered is to limit diffusive processes spread over the network, for example mitigating pandemic disease or computer virus spread. A number of problem formulations have been proposed that aim to solve such problems based on desired network characteristics, such as maintaining the largest network component after node removal. The recently formulated critical node detection problem aims to remove a small subset of vertices from the network such that the residual network has minimum pairwise connectivity. Unfortunately, the problem is NP-hard and also the number of constraints is cubic in number of vertices, making very large scale problems impossible to solve with traditional mathematical programming techniques. Even many approximation algorithm strategies such as dynamic programming, evolutionary algorithms, etc. all are unusable for networks that contain thousands to millions of vertices. A computationally efficient and simple approach is required in such circumstances, but none currently exist. In this thesis, such an algorithm is proposed. The methodology is based on a depth-first search traversal of the network, and a specially designed ranking function that considers information local to each vertex. Due to the variety of network structures, a number of characteristics must be taken into consideration and combined into a single rank that measures the utility of removing each vertex. Since removing a vertex in sequential fashion impacts the network structure, an efficient post-processing algorithm is also proposed to quickly re-rank vertices. Experiments on a range of common complex network models with varying number of vertices are considered, in addition to real world networks. The proposed algorithm, DFSH, is shown to be highly competitive and often outperforms existing strategies such as Google PageRank for minimizing pairwise connectivity.

Psychopathy and the Ability to Detect Cheaters: An Evolutionary Perspective

This research examined psychopathy as an evolutionary adaptation that involves cheating and deception. I theorized that this strategy should be associated with certain abilities. This research examined the association between psychopathic traits and the ability to detect cheaters, altruism, deception, and psychopathic traits. Results indicated that psychopathic traits were not significantly associated with the ability to detect cheaters or altruism. Results indicated that high Factor 1 psychopathy scores, and low Factor 2 psychopathy scores, were indicative of higher ratings of deception when viewing deceptive videos. Conversely, when viewing truthful videos, Factor 1 was a significant predictor of higher ratings of deception. Finally, our results indicated that total psychopathy scores were associated the ability to identify psychopathic traits in others. Taken together the results provide mixed support for the evolutionary perspective of psychopathy.