Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.

Comparing the alternative utilization method for coconut in Nzema, Ghana

There are many opportunities to utilise coconut in Nzema to support farmers. Coconut oil that is mainly used for food preparation in Nzema can be utilized as fuel to support overcoming of the energy crisis in the Ghana. Coconut oil in Nzema is not used in both transportation and electricity generation. A few of the waste husk and shell are mainly used as fuel in homes for heating but greater amount is left to rot or burn the coconut plantation. In addition, some portion of the granulated coconut kernel is sometime used as feed for piggery feed and the rest of the granulated kernel are left as waste on the oil processing site. In this thesis, the author identified alternative utilization of cocoanut, for instance the use of coconut husk and shell for charcoal production, and the use of coconut trunks as construction materials. It is envisaged that exploring these alternatives will not only reduce carbon emission in the country but will also contribute significantly to the sustainability of the local agro-industry.

Impact of regeneration method on stand structure prior to first thinning

Summary

A method for using random parameters in analyzing permanent sample plots

Summary

Phosphorus saturation of Finnish soils : evaluating an easy oxalate extraction method

Selostus: Alumiini- ja rautaoksidien fosforikyllästysasteen arvioiminen suomalaisista peltomaista

A new heuristic method for solving spatially constrained forest planning problems based on mitigation of infeasibilities radiating outward from a forced choice

[Abstract]

Evaluation of the multicriteria approval method for timber-harvesting group decision support

Abstract

Effective notch method assessment of misalignment for fatigue loaded welds

 The objective of this thesis work was to assess axial misalignment in fatigue loaded welds using the effective notch method. As a result, the fatigue behaviour of non-load carrying cruciform fillet welded joint under cyclic tensile loading has been studied. Various degrees of axial misalignment have been found in one series of non-load carrying cruciform fillet welded joints used in a laboratory investigation. As a result, it was important to carry out a comprehensive investigation since axial misalignment forms part of thequality of fatigue loaded structure and can reduce the fatigue strength. To extend the study, the correlation between fatigue strength and stress ratio, as well as stress concentration factor, were also studied. Moreover, a closer investigation of place of crack initiation and its dependence on weld sequence and imperfections of test specimen (angular distortion) was studied. For the fatigue class calculations, FEM (finite element method) and the effectivenotch approach are used. The addressed variable is the axial misalignment whichis introduce by modeling the entire joint. Fracture mechanics based calculations are also used and quantitatively compared with effective notch and experimental results.

Optimisation of data analysis method for a fluidized bed combustion test rig

Nowadays the used fuel variety in power boilers is widening and new boiler constructions and running models have to be developed. This research and development is done in small pilot plants where more faster analyse about the boiler mass and heat balance is needed to be able to find and do the right decisions already during the test run. The barrier on determining boiler balance during test runs is the long process of chemical analyses of collected input and outputmatter samples. The present work is concentrating on finding a way to determinethe boiler balance without chemical analyses and optimise the test rig to get the best possible accuracy for heat and mass balance of the boiler. The purpose of this work was to create an automatic boiler balance calculation method for 4 MW CFB/BFB pilot boiler of Kvaerner Pulping Oy located in Messukylä in Tampere. The calculation was created in the data management computer of pilot plants automation system. The calculation is made in Microsoft Excel environment, which gives a good base and functions for handling large databases and calculations without any delicate programming. The automation system in pilot plant was reconstructed und updated by Metso Automation Oy during year 2001 and the new system MetsoDNA has good data management properties, which is necessary for big calculations as boiler balance calculation. Two possible methods for calculating boiler balance during test run were found. Either the fuel flow is determined, which is usedto calculate the boiler's mass balance, or the unburned carbon loss is estimated and the mass balance of the boiler is calculated on the basis of boiler's heat balance. Both of the methods have their own weaknesses, so they were constructed parallel in the calculation and the decision of the used method was left to user. User also needs to define the used fuels and some solid mass flowsthat aren't measured automatically by the automation system. With sensitivity analysis was found that the most essential values for accurate boiler balance determination are flue gas oxygen content, the boiler's measured heat output and lower heating value of the fuel. The theoretical part of this work concentrates in the error management of these measurements and analyses and on measurement accuracy and boiler balance calculation in theory. The empirical part of this work concentrates on the creation of the balance calculation for the boiler in issue and on describing the work environment.

A support system for evaluating a suitable joining method in the production of sheet metal goods

This report illustrates a comparative study of various joining methods involved in sheet metal production. In this report it shows the selection of joining methods, which includes comparing the advantages and disadvantages of a method over the other ones and choosing the best method for joining. On the basis of various joining process from references, a table is generated containing set of criterion that helps in evaluation of various sheet metal joining processes and in selecting the most suitable process for a particular product. Three products are selected and a comprehensive study of the joining methods is analyzed with the help of various parameters. The table thus is the main part of the analysis process of this study and can be advanced with the beneficial results. It helps in a better and easy understanding and comparing the various methods, which provides the foundation of this study and analysis. The suitability of the joining method for various types of cases of different sheet metal products can be tested with the help of this table. The sections of the created table display the requirements of manufacturing. The important factor has been considered and given focus in the table, as how the usage of these parameters is important in percentages according to particular or individual case. The analysis of the methods can be extended or altered by changing the parameters according to the constraint. The use of this table is demonstrated by pertaining the cases from sheet metal production.