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.

Hyperspectral Mineral Identification using SVM and SOM

Remote sensing techniques involving hyperspectral imagery have applications in a number of sciences that study some aspects of the surface of the planet. The analysis of hyperspectral images is complex because of the large amount of information involved and the noise within that data. Investigating images with regard to identify minerals, rocks, vegetation and other materials is an application of hyperspectral remote sensing in the earth sciences. This thesis evaluates the performance of two classification and clustering techniques on hyperspectral images for mineral identification. Support Vector Machines (SVM) and Self-Organizing Maps (SOM) are applied as classification and clustering techniques, respectively. Principal Component Analysis (PCA) is used to prepare the data to be analyzed. The purpose of using PCA is to reduce the amount of data that needs to be processed by identifying the most important components within the data. A well-studied dataset from Cuprite, Nevada and a dataset of more complex data from Baffin Island were used to assess the performance of these techniques. The main goal of this research study is to evaluate the advantage of training a classifier based on a small amount of data compared to an unsupervised method. Determining the effect of feature extraction on the accuracy of the clustering and classification method is another goal of this research. This thesis concludes that using PCA increases the learning accuracy, and especially so in classification. SVM classifies Cuprite data with a high precision and the SOM challenges SVM on datasets with high level of noise (like Baffin Island).

DNA polymerase activity in trophoblast giant cells in the mouse /

In the developing mouse embryo, the diploid trophectoderm is known to undergo a diploid to giant cell transformation. These cells arise by a process of endoreduplication, characterized by replication of the entire genome without subsequent mitosis or cell division, leading to polyploidy and the formation of giant nuclei. Studies of 13.5 day rat trophoblast derived from the parietal yolk sac have indicated a relatively low rate of DNA polymerase a activity, the noinnal eukaryotic replicase, in comparison to that of DNA polymerase g. These results have suggested that endoreduplication in trophoblast giant cells may not employ the normal replicase enzyme, DNA polymerase a. In order to determine whether a 'switch' from DNA polymerase to DNA polymerase is a necessary concomitant of the diploid to giant cell transformation, two distinct populations of trophoblast giant cells, the primary giant cell derived from the mural trophectoderm and the secondary giant cell derived from the polar trophoectoderm were used. These two populations of trophoblast giant cells can be obtained from the tissue outgrowths of 3.5da blastocysts and the extraembryonic ectoderm (EX) and ectoplacental cone (EPC) of 7.5 day embryos respectively. Tissue outgrowths were treated with aphidicolin, a specific reversible inhibitor of eukaryotic DNA polymerase a, on various days after explantation. The effect of aphidicolin treatment was assessed both qualitatively, using autoradiography and quantitatively by scintillation counting and Feulgen staining. 3 DNA synthesis was measured in control and treated cultures after a Hthymidine pulse. Scintillation counts of the embryo proper revealed that DNA synthesis was consistently inhibited by greater than 907. in the presence of aphidicolin. Inhibition of DNA synthesis in the EX and EPC varied between 81-957. and 82-987. respectively, indicating that most DNA synthesis was mediated by DNA polymerase a, but that a small but significant amount of residual synthesis was indicated. A qualitative approach was then applied to determine whether the apparent residual DNA synthesis was restricted to a subpopulation of giant cells or whether all giant cells displayed a low level of DNA synthesis. Autoradiographs of the ICM of blastocysts and the embryo proper of 7.5da embryos, which acted as diploid control population, was completely inhibited regardless of duration in explant culture. In contrast, primary trophoblast giant cells derived from blastocysts and secondary giant cells derived from the EX and EPC were observed to possess some heavily labelled cells after aphidicolin treatment. These results suggest that although DNA polymerase a is the primary replicating enzyme responsible for endoreduplication in mouse trophoblast giant cells, some nonactivity is also observed. A DNA polymerase assay employing tissue lysates of outgrown 7.5da embryo, EX and EPC tissues was used to attempt to confirm the presence of higher nonactivity in tissues possessing trophoblast giant cells. Employing a series of inhibitors of DNA polymerases, it would appear that DNA polymerase a is the major polymerase active in all tissues of the 7.5da mouse embryo. The nature of the putative residual DNA synthetic activity could not be unequivically determined in this study. Therefore, these results suggest that both primary and secondary trophoblast giant cells possess and use DNA polymerase a in endoreduplicative DNA synthesis. It would appear that the high levels of DNA polymerase g activity reported in trophoblast tissue derived from the 13.5 da rat yolk sac was not a general feature of all endoreduplication.

Diatoms as indicators of lake trophic status

Relationships between surface sediment diatom assemblages and lake trophic status were studied in 50 Canadian Precambrian Shield lakes in the Muskoka-Haliburton and southern Ontario regions. The purpose of this study was to develop mathematical regression models to infer lake trophic status from diatom assemblage data. To achieve this goal, however, additional investigations dealing with the evaluation of lake trophic status and the autecological features of key diatom species were carried out. Because a unifying index and classification for lake trophic status was not available, a new multiple index was developed in this study, by the computation of the physical, chemical and biological data from 85 south Ontario lakes. By using the new trophic parameter, the lake trophic level (TL) was determined: TL = 1.37 In[1 +(TP x Chl-a / SD)], where, TP=total phosphorus, Chl-a=chlorophyll-a and SD=Secchi depth. The boundaries between 7 lake trophic categories (Ultra-oligotrophic lakes: 0-0.24; Oligotrophic lakes: 0.241-1.8; Oligomesotrophic lakes: 1.813.0; Mesotrophic lakes: 3.01-4.20; Mesoeutrophic lakes: 4.21-5.4; Eutrophic lakes: 5.41-10 and Hyper-eutrophic lakes: above 10) were established. The new trophic parameter was more convenient for management of water quality, communication to the public and comparison with other lake trophic status indices than many of the previously published indices because the TL index attempts to Increase understanding of the characteristics of lakes and their comprehensive trophic states. It is more reasonable and clear for a unifying determination of true trophic states of lakes. Diatom specIes autecology analysis was central to this thesis. However, the autecological relationship of diatom species and lake trophic status had not previously been well documented. Based on the investigation of the diatom composition and variety of species abundance in 30 study lakes, the distribution optima of diatom species were determined. These determinations were based on a quantitative method called "weighted average" (Charles 1985). On this basis, the diatom species were classified into five trophic categories (oligotrophic, oligomesotrophic, mesotrophic, mesoeutrophic and eutrophic species groups). The resulting diatom trophic status autecological features were used in the regressIon analysis between diatom assemblages and lake trophic status. When the TL trophic level values of the 30 lakes were regressed against their fi ve corresponding diatom trophic groups, the two mathematical equations for expressing the assumed linear relationship between the diatom assemblages composition were determined by (1) uSIng a single regression technique: Trophic level of lake (TL) = 2.643 - 7.575 log (Index D) (r = 0.88 r2 = 0.77 P = 0.0001; n = 30) Where, Index D = (0% + OM% + M%)/(E% + ME% + M%); 4 (2) uSIng a' multiple regressIon technique: TL=4.285-0.076 0%- 0.055 OM% - 0.026 M% + 0.033 ME% + 0.065 E% (r=0.89, r2=0.792, P=O.OOOl, n=30) There was a significant correlation between measured and diatom inferred trophic levels both by single and multiple regressIon methods (P < 0.0001, n=20), when both models were applied to another 20 test lakes. Their correlation coefficients (r2 ) were also statistically significant (r2 >0.68, n=20). As such, the two transfer function models between diatoms and lake trophic status were validated. The two models obtained as noted above were developed using one group of lakes and then tested using an entirely different group of lakes. This study indicated that diatom assemblages are sensitive to lake trophic status. As indicators of lake trophic status, diatoms are especially useful in situations where no local trophic information is available and in studies of the paleotrophic history of lakes. Diatom autecological information was used to develop a theory assessing water quality and lake trophic status.

Upper Lakes Shipping Ltd. Vessels Scrapbook, 1964-1967, no date

The origins of Upper Lakes Shipping can be traced back to 1931, when Gordon Leitch, the general manager of Toronto elevators, sent Captain Bruce Angus to scout for potential ships that could be used to transport grain. The ship Sarnian was purchased as a result. It soon became apparent that more ships would be needed by the business, and Leitch subsequently partnered with James Norris of the Norris Grain Company, in order to transport their products more economically. The Upper Lakes and St. Lawrence Transportation Company (which later become known as Upper Lakes Shipping), was incorporated in 1932. In 1956, the company bought Port Weller Dry docks. At this point, the company began constructing new vessels designed to meet specific freight and classification needs.