\relax \citation{Hart68a} \citation{Frank+Asuncion:2010} \citation{Hart67} \citation{Bezdek2000,Bezdek2001,Kim2003} \@writefile{toc}{\contentsline {chapter}{\numberline {1}Introduction}{1}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{introduction}{{1}{1}} \citation{jalali08} \citation{Frank+Asuncion:2010} \citation{fastmap} \citation{01cod} \citation{01cod} \@writefile{toc}{\contentsline {section}{\numberline {1.1}Statement of Thesis}{4}} \@writefile{toc}{\contentsline {section}{\numberline {1.2}Contributions of this Thesis}{4}} \@writefile{toc}{\contentsline {section}{\numberline {1.3}Structure of this Thesis}{4}} \citation{Hart68a} \citation{Devi2002,Bezdek2001,Chang74,Dasarathy94,Li2009,Bezdek98,Cano2005,Garain2008,Veenman2005,GarcĂ­a2008,Kohonen98,Kohonen90} \@writefile{toc}{\contentsline {chapter}{\numberline {2}Background and Related Work}{5}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{related}{{2}{5}} \citation{Hart68a} \citation{Chang74} \citation{Hart68a} \@writefile{toc}{\contentsline {section}{\numberline {2.1}Prototype Learning for Nearest Neighbor Classifiers}{6}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.1.1}Instance Selection}{6}} \@writefile{toc}{\contentsline {subsubsection}{Condensed Nearest Neighbor (CNN)}{6}} \citation{Gates72} \@writefile{lof}{\contentsline {figure}{\numberline {2.1}{\ignorespaces Pseudo-code for CNN}}{7}} \newlabel{fig:cnn}{{2.1}{7}} \citation{Hart68a} \citation{Dasarathy94} \@writefile{toc}{\contentsline {subsubsection}{Reduced Nearest Neighbor (RNN)}{8}} \@writefile{lof}{\contentsline {figure}{\numberline {2.2}{\ignorespaces Pseudo-code for RNN}}{8}} \newlabel{fig:rnn}{{2.2}{8}} \@writefile{toc}{\contentsline {subsubsection}{Minimal Consistent Set (MCS)}{8}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.1.2}Instance Abstraction}{9}} \@writefile{toc}{\contentsline {subsubsection}{Chang}{9}} \@writefile{lof}{\contentsline {figure}{\numberline {2.3}{\ignorespaces Pseudo-code for MCS}}{10}} \newlabel{fig:mcs1}{{2.3}{10}} \@writefile{toc}{\contentsline {section}{\numberline {2.2}Evaluation of Prototype Learning Schemes}{10}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.2.1}Storage Reduction}{10}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.2.2}Speed Increase}{10}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.2.3}Generalization Accuracy}{10}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.2.4}Noise Tolerance}{10}} \@writefile{toc}{\contentsline {subsection}{\numberline {2.2.5}Probability of Detection and False Alarm}{10}} \@writefile{lof}{\contentsline {figure}{\numberline {2.4}{\ignorespaces Chang's algorithm for finding prototypes}}{11}} \newlabel{fig:chang}{{2.4}{11}} \citation{jalali08} \@writefile{toc}{\contentsline {chapter}{\numberline {3}CLIFF: Tool for Instance Selection}{12}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{cliff}{{3}{12}} \citation{jalali08} \@writefile{toc}{\contentsline {section}{\numberline {3.1}CLIFF}{13}} \newlabel{section:selector}{{3.1}{13}} \newlabel{eq:one1}{{3.1}{13}} \newlabel{eq:one}{{3.2}{13}} \@writefile{lof}{\contentsline {figure}{\numberline {3.1}{\ignorespaces Pseudo code for Support Based Bayesian Ranking algorithm}}{14}} \newlabel{fig:sbbra}{{3.1}{14}} \@writefile{toc}{\contentsline {section}{\numberline {3.2}CLIFF: A Simple Example}{15}} \@writefile{lof}{\contentsline {figure}{\numberline {3.2}{\ignorespaces A log of some golf-playing behavior}}{15}} \newlabel{fig:golf}{{3.2}{15}} \@writefile{lof}{\contentsline {figure}{\numberline {3.3}{\ignorespaces Finding the rank of $sunny$}}{16}} \newlabel{fig:rank1}{{3.3}{16}} \@writefile{toc}{\contentsline {section}{\numberline {3.3}CLIFF: Time Complexity}{16}} \newlabel{section:time}{{3.3}{16}} \citation{Frank+Asuncion:2010} \citation{fastmap} \@writefile{toc}{\contentsline {chapter}{\numberline {4}CLIFF Assessment}{17}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{assess}{{4}{17}} \@writefile{toc}{\contentsline {section}{\numberline {4.1}Data and Preprocessing Tools}{17}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.1.1}Data Set Characteristics}{17}} \@writefile{lof}{\contentsline {figure}{\numberline {4.1}{\ignorespaces Data Set Characteristics}}{18}} \newlabel{fig:info}{{4.1}{18}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.1.2}Pre-processing tools for Dimensionality Reduction}{18}} \@writefile{toc}{\contentsline {subsubsection}{FastMap}{18}} \citation{Hart68a} \citation{Dasarathy94} \citation{lot2010} \@writefile{lof}{\contentsline {figure}{\numberline {4.2}{\ignorespaces Data Set Characteristics}}{19}} \newlabel{fig:work}{{4.2}{19}} \@writefile{toc}{\contentsline {subsubsection}{Feature Subset Selection (FSS)}{19}} \@writefile{lof}{\contentsline {figure}{\numberline {4.3}{\ignorespaces Choosing the best number of features for each data set. The best choice will have a high pd along with a low pf}}{20}} \newlabel{fig:numf}{{4.3}{20}} \@writefile{lof}{\contentsline {figure}{\numberline {4.4}{\ignorespaces Example of using the cosine law to find the position of $Oi$ in the dimension $k$}}{20}} \newlabel{fig:fm1}{{4.4}{20}} \citation{Frank+Asuncion:2010} \@writefile{lof}{\contentsline {figure}{\numberline {4.5}{\ignorespaces Projects of points $O_i$ and $O_j$ onto the hyper-plane perpendicular to the line $O_a$$O_b$}}{21}} \newlabel{fig:fm2}{{4.5}{21}} \@writefile{lof}{\contentsline {figure}{\numberline {4.6}{\ignorespaces Data Set Characteristics}}{22}} \newlabel{fig:info}{{4.6}{22}} \@writefile{toc}{\contentsline {section}{\numberline {4.2}CLIFF Assessment on Standard Data Sets}{22}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.2.1}Data}{22}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.2.2}Experimental Method}{22}} \newlabel{section:brit}{{4.2.2}{22}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.2.3}Experiment ##1: Is CLIFF viable as a Prototype Learning Scheme for NNC?}{23}} \@writefile{toc}{\contentsline {subsection}{\numberline {4.2.4}Experiment ##2: How well does CLIFF handle the presence of noise?}{23}} \@writefile{lof}{\contentsline {figure}{\numberline {4.7}{\ignorespaces Pseudo code for Experiment ##1}}{24}} \newlabel{fig:knnexp1}{{4.7}{24}} \@writefile{lof}{\contentsline {figure}{\numberline {4.8}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -iris}}{25}} \newlabel{fig:results100}{{4.8}{25}} \@writefile{lof}{\contentsline {figure}{\numberline {4.9}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -breastcancer}}{25}} \newlabel{fig:results101}{{4.9}{25}} \@writefile{lof}{\contentsline {figure}{\numberline {4.10}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -mam}}{26}} \newlabel{fig:results102}{{4.10}{26}} \@writefile{lof}{\contentsline {figure}{\numberline {4.11}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -hc}}{26}} \newlabel{fig:results103}{{4.11}{26}} \@writefile{lof}{\contentsline {figure}{\numberline {4.12}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -hh}}{27}} \newlabel{fig:results104}{{4.12}{27}} \@writefile{lof}{\contentsline {figure}{\numberline {4.13}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -hs}}{27}} \newlabel{fig:results105}{{4.13}{27}} \@writefile{lof}{\contentsline {figure}{\numberline {4.14}{\ignorespaces Probability of Detection (PD) and Probability of False Alarm (PF)results -hv}}{28}} \newlabel{fig:results106}{{4.14}{28}} \newlabel{fig:charts1}{{4.15}{29}} \newlabel{fig:charts2}{{4.16}{30}} \citation{Walsh94} \@writefile{toc}{\contentsline {chapter}{\numberline {5}Case Study: Solving the Problem of Brittleness in Forensic Models}{31}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{forensics}{{5}{31}} \@writefile{toc}{\contentsline {section}{\numberline {5.1}Introduction}{31}} \citation{09NAS} \citation{09NAS} \citation{Seheult78} \citation{Grove80} \@writefile{toc}{\contentsline {section}{\numberline {5.2}Visualization of Brittleness}{33}} \@writefile{toc}{\contentsline {section}{\numberline {5.3}Glass Forensic Models}{33}} \citation{Evett94} \citation{Walsh94} \citation{Seheult78} \citation{77Lindley} \@writefile{toc}{\contentsline {subsection}{\numberline {5.3.1}Seheult 1978}{34}} \newlabel{subsection:seh}{{5.3.1}{34}} \newlabel{eq:lin1}{{5.1}{34}} \newlabel{eq:lin2}{{5.2}{34}} \citation{Evett94} \@writefile{toc}{\contentsline {subsection}{\numberline {5.3.2}Grove 1980}{35}} \newlabel{subsection:gro}{{5.3.2}{35}} \newlabel{eq:gro2}{{5.3}{35}} \newlabel{eq:gro4}{{5.5}{35}} \@writefile{toc}{\contentsline {subsection}{\numberline {5.3.3}Evett 1995}{36}} \newlabel{subsection:eve}{{5.3.3}{36}} \citation{Walsh94} \@writefile{toc}{\contentsline {subsection}{\numberline {5.3.4}Walsh 1996}{37}} \newlabel{subsection:wal}{{5.3.4}{37}} \@writefile{toc}{\contentsline {section}{\numberline {5.4}Visualization of Brittleness in Models}{38}} \@writefile{lof}{\contentsline {figure}{\numberline {5.1}{\ignorespaces Visualization of four(4) glass forensic models}}{39}} \newlabel{fig:models}{{5.1}{39}} \citation{09NAS} \citation{Seheult78,Evett84,Evett90,Evett94,Walsh94} \citation{Seheult78,Grove80,Walsh94} \citation{Karslake09} \citation{Karslake09} \citation{Karslake09} \@writefile{toc}{\contentsline {section}{\numberline {5.5}Introduction}{40}} \citation{Karslake09} \citation{joll02} \@writefile{lof}{\contentsline {figure}{\numberline {5.2}{\ignorespaces Proposed procedure for the forensic evaluation of data}}{42}} \newlabel{fig:process}{{5.2}{42}} \@writefile{toc}{\contentsline {section}{\numberline {5.6}Dimensionality Reduction}{42}} \@writefile{toc}{\contentsline {subsection}{\numberline {5.6.1}Principal Component Analysis}{42}} \@writefile{lof}{\contentsline {figure}{\numberline {5.3}{\ignorespaces PCA for iris data set}}{43}} \newlabel{fig:iris}{{5.3}{43}} \citation{Karslake09} \citation{pca} \citation{Karslake09} \citation{fastmap} \citation{knn} \@writefile{toc}{\contentsline {section}{\numberline {5.7}Clustering}{45}} \@writefile{toc}{\contentsline {section}{\numberline {5.8}Classification with KNN}{45}} \newlabel{subsection:knn}{{5.8}{45}} \@writefile{lof}{\contentsline {figure}{\numberline {5.4}{\ignorespaces Probability of detection (pd) and Probability of False alarms (pf) using fixed values for dimensions and fixed k values for k-nearest neighbor}}{46}} \newlabel{fig:exp1}{{5.4}{46}} \@writefile{lof}{\contentsline {figure}{\numberline {5.5}{\ignorespaces Pseudo code for K-means}}{47}} \newlabel{fig:kmeans}{{5.5}{47}} \@writefile{toc}{\contentsline {section}{\numberline {5.9}The Brittleness Measure}{47}} \newlabel{subsection:bm}{{5.9}{47}} \citation{Karslake09} \citation{Karslake09} \citation{Karslake09} \newlabel{eq:bm}{{5.9}{48}} \@writefile{toc}{\contentsline {section}{\numberline {5.10}Data Set and Experimental Method}{48}} \newlabel{section:brit}{{5.10}{48}} \@writefile{toc}{\contentsline {section}{\numberline {5.11}Experiment 1: KNN as a forensic model?}{49}} \@writefile{toc}{\contentsline {subsection}{\numberline {5.11.1}Results from Experiment 1}{49}} \@writefile{lof}{\contentsline {figure}{\numberline {5.6}{\ignorespaces Pseudo code for Experiment 1}}{50}} \newlabel{fig:knnexp1}{{5.6}{50}} \@writefile{toc}{\contentsline {section}{\numberline {5.12}Experiment 2: Can brittleness be reduced?}{50}} \@writefile{lof}{\contentsline {figure}{\numberline {5.7}{\ignorespaces Results for Experiment 1 for the 4 data sets distinguished by the number of clusters. Here for the upper and lower tables n=4 is used while r=1 is used for the upper table and r=2 for the lower table.}}{51}} \newlabel{fig:result1}{{5.7}{51}} \@writefile{lof}{\contentsline {figure}{\numberline {5.8}{\ignorespaces Pseudo code for Experiment 2}}{52}} \newlabel{fig:knnexp2}{{5.8}{52}} \@writefile{toc}{\contentsline {subsection}{\numberline {5.12.1}Results from Experiment 2}{52}} \citation{09Zadora,09aZadora,06Aitken,04Aitken,02Koons,99Koons} \@writefile{lof}{\contentsline {figure}{\numberline {5.9}{\ignorespaces Results for Experiment 2 for the 4 data sets distinguished by the number of clusters. Here for the upper and lower tables n=4 is used while r=1 is used for the upper table and r=2 for the lower table.}}{54}} \newlabel{fig:result2}{{5.9}{54}} \@writefile{lof}{\contentsline {figure}{\numberline {5.10}{\ignorespaces Position of values in the 'before' and 'after' population with data set at 3, 5, 10 and 20 clusters. The first row shows the results for r=1 while the second row shows the results for r=2}}{55}} \newlabel{fig:dist3}{{5.10}{55}} \@writefile{lof}{\contentsline {figure}{\numberline {5.11}{\ignorespaces Results for Experiment 2 of before and after results. -1 indicates that the after is better than before}}{55}} \newlabel{fig:result3}{{5.11}{55}} \citation{09NAS} \citation{09NAS} \citation{Karslake09} \@writefile{toc}{\contentsline {chapter}{\numberline {6}Conclusion}{56}} \@writefile{lof}{\addvspace {10\p@ }} \@writefile{lot}{\addvspace {10\p@ }} \newlabel{conclusion}{{6}{56}} \bibstyle{plain} \bibdata{thesis-v2} \bibcite{04Aitken}{1} \bibcite{06Aitken}{2} \bibcite{Bezdek2000}{3} \bibcite{Bezdek2001}{4} \bibcite{pca}{5} \bibcite{Chang74}{6} \bibcite{Hart67}{7} \bibcite{Dasarathy94}{8} \bibcite{knn}{9} \bibcite{Evett84}{10} \bibcite{Evett90}{11} \bibcite{Evett94}{12} \bibcite{fastmap}{13} \bibcite{Frank+Asuncion:2010}{14} \bibcite{Grove80}{15} \bibcite{Hart68a}{16} \bibcite{jalali08}{17} \bibcite{joll02}{18} \bibcite{Karslake09}{19} \bibcite{Kim2003}{20} \bibcite{02Koons}{21} \bibcite{99Koons}{22} \bibcite{01cod}{23} \bibcite{77Lindley}{24} \bibcite{09NAS}{25} \bibcite{Seheult78}{26} \bibcite{Walsh94}{27} \bibcite{09aZadora}{28} \bibcite{09Zadora}{29}