%PDF-1.4
%
1 0 obj
<< /S /GoTo /D (chapter.1) >>
endobj
4 0 obj
(Introduction)
endobj
5 0 obj
<< /S /GoTo /D (section.1.1) >>
endobj
8 0 obj
(Contributions)
endobj
9 0 obj
<< /S /GoTo /D (section.1.2) >>
endobj
12 0 obj
(Contents)
endobj
13 0 obj
<< /S /GoTo /D (chapter.2) >>
endobj
16 0 obj
(Inductive Rule Learning)
endobj
17 0 obj
<< /S /GoTo /D (section.2.1) >>
endobj
20 0 obj
(Foundations of Machine Learning and Inductive Rule Learning)
endobj
21 0 obj
<< /S /GoTo /D (subsection.2.1.1) >>
endobj
24 0 obj
(Classification)
endobj
25 0 obj
<< /S /GoTo /D (subsection.2.1.2) >>
endobj
28 0 obj
(Regression)
endobj
29 0 obj
<< /S /GoTo /D (section.2.2) >>
endobj
32 0 obj
(Separate-and-Conquer Rule Learning)
endobj
33 0 obj
<< /S /GoTo /D (subsection.2.2.1) >>
endobj
36 0 obj
(Introduction of Rules)
endobj
37 0 obj
<< /S /GoTo /D (subsection.2.2.2) >>
endobj
40 0 obj
(A straight-forward Separate-and-Conquer Algorithm)
endobj
41 0 obj
<< /S /GoTo /D (subsection.2.2.3) >>
endobj
44 0 obj
(Searching for good Rules)
endobj
45 0 obj
<< /S /GoTo /D (subsection.2.2.4) >>
endobj
48 0 obj
(Discussion of the straight-forward Algorithm)
endobj
49 0 obj
<< /S /GoTo /D (subsection.2.2.5) >>
endobj
52 0 obj
(Characterizing Separate-and-Conquer Algorithms)
endobj
53 0 obj
<< /S /GoTo /D (section.2.3) >>
endobj
56 0 obj
(Handling Multi-Class Problems)
endobj
57 0 obj
<< /S /GoTo /D (subsection.2.3.1) >>
endobj
60 0 obj
(\(Unordered\) 1-vs-all Class Binarization)
endobj
61 0 obj
<< /S /GoTo /D (subsection.2.3.2) >>
endobj
64 0 obj
(Ordered 1-vs-all Class Binarization)
endobj
65 0 obj
<< /S /GoTo /D (subsection.2.3.3) >>
endobj
68 0 obj
(Pairwise Class Binarization)
endobj
69 0 obj
<< /S /GoTo /D (subsection.2.3.4) >>
endobj
72 0 obj
(Ordered and Unordered Lists of Rules)
endobj
73 0 obj
<< /S /GoTo /D (section.2.4) >>
endobj
76 0 obj
(Overfitting Avoidance)
endobj
77 0 obj
<< /S /GoTo /D (section.2.5) >>
endobj
80 0 obj
(Visualization with Coverage Space Isometrics)
endobj
81 0 obj
<< /S /GoTo /D (section.2.6) >>
endobj
84 0 obj
(Rule Learning Heuristics)
endobj
85 0 obj
<< /S /GoTo /D (subsection.2.6.1) >>
endobj
88 0 obj
(Basic Heuristics)
endobj
89 0 obj
<< /S /GoTo /D (subsection.2.6.2) >>
endobj
92 0 obj
(Composite Heuristics)
endobj
93 0 obj
<< /S /GoTo /D (subsection.2.6.3) >>
endobj
96 0 obj
(Parametrized Heuristics)
endobj
97 0 obj
<< /S /GoTo /D (subsection.2.6.4) >>
endobj
100 0 obj
(Gain-Heuristics)
endobj
101 0 obj
<< /S /GoTo /D (section.2.7) >>
endobj
104 0 obj
(Evaluation Methods for Rule Learning Algorithms)
endobj
105 0 obj
<< /S /GoTo /D (subsection.2.7.1) >>
endobj
108 0 obj
(Cross-Validation)
endobj
109 0 obj
<< /S /GoTo /D (subsection.2.7.2) >>
endobj
112 0 obj
(Theory size)
endobj
113 0 obj
<< /S /GoTo /D (subsection.2.7.3) >>
endobj
116 0 obj
(Evaluation of Regression Rule Learning Algorithms)
endobj
117 0 obj
<< /S /GoTo /D (subsection.2.7.4) >>
endobj
120 0 obj
(Evaluation of Classification Rule Learning Algorithms)
endobj
121 0 obj
<< /S /GoTo /D (subsection.2.7.5) >>
endobj
124 0 obj
(Evaluation of Performance Rankings)
endobj
125 0 obj
<< /S /GoTo /D (subsection.2.7.6) >>
endobj
128 0 obj
(Pairwise Comparisons using the Sign Test)
endobj
129 0 obj
<< /S /GoTo /D (chapter.3) >>
endobj
132 0 obj
(The SeCo-Framework for Rule Learning)
endobj
133 0 obj
<< /S /GoTo /D (section.3.1) >>
endobj
136 0 obj
(Separate-and-conquer Rule Learning in the SeCo-Framework)
endobj
137 0 obj
<< /S /GoTo /D (section.3.2) >>
endobj
140 0 obj
(Unifying Rule Learners in the SeCo-Framework)
endobj
141 0 obj
<< /S /GoTo /D (subsection.3.2.1) >>
endobj
144 0 obj
(Fixed Properties of the SeCo-Framework )
endobj
145 0 obj
<< /S /GoTo /D (section.3.3) >>
endobj
148 0 obj
(Architecture of the SeCo-Framework)
endobj
149 0 obj
<< /S /GoTo /D (section.3.4) >>
endobj
152 0 obj
(Configurable Objects in the Framework)
endobj
153 0 obj
<< /S /GoTo /D (subsection.3.4.1) >>
endobj
156 0 obj
(Binarization in the SeCo-Framework)
endobj
157 0 obj
<< /S /GoTo /D (section.3.5) >>
endobj
160 0 obj
(Example Configurations)
endobj
161 0 obj
<< /S /GoTo /D (subsection.3.5.1) >>
endobj
164 0 obj
(CN2)
endobj
165 0 obj
<< /S /GoTo /D (subsection.3.5.2) >>
endobj
168 0 obj
(AQ)
endobj
169 0 obj
<< /S /GoTo /D (subsection.3.5.3) >>
endobj
172 0 obj
(Ripper)
endobj
173 0 obj
<< /S /GoTo /D (subsection.3.5.4) >>
endobj
176 0 obj
(SimpleSeCo)
endobj
177 0 obj
<< /S /GoTo /D (section.3.6) >>
endobj
180 0 obj
(Evaluation Package)
endobj
181 0 obj
<< /S /GoTo /D (section.3.7) >>
endobj
184 0 obj
(Related Work)
endobj
185 0 obj
<< /S /GoTo /D (section.3.8) >>
endobj
188 0 obj
(Summary)
endobj
189 0 obj
<< /S /GoTo /D (chapter.4) >>
endobj
192 0 obj
(Heuristics for Classification)
endobj
193 0 obj
<< /S /GoTo /D (section.4.1) >>
endobj
196 0 obj
(Experimental Setup)
endobj
197 0 obj
<< /S /GoTo /D (subsection.4.1.1) >>
endobj
200 0 obj
(The Datasets)
endobj
201 0 obj
<< /S /GoTo /D (section.4.2) >>
endobj
204 0 obj
(Optimization of Parametrized Heuristics)
endobj
205 0 obj
<< /S /GoTo /D (subsection.4.2.1) >>
endobj
208 0 obj
(Search Strategy)
endobj
209 0 obj
<< /S /GoTo /D (subsection.4.2.2) >>
endobj
212 0 obj
(Optimal Parameters for the Five Heuristics)
endobj
213 0 obj
<< /S /GoTo /D (subsection.4.2.3) >>
endobj
216 0 obj
(Experimental Results of the Tuned Heuristics)
endobj
217 0 obj
<< /S /GoTo /D (subsection.4.2.4) >>
endobj
220 0 obj
(Interpretation of the Learned Heuristics)
endobj
221 0 obj
<< /S /GoTo /D (section.4.3) >>
endobj
224 0 obj
(Meta-Learning of Rule Learning Heuristics)
endobj
225 0 obj
<< /S /GoTo /D (subsection.4.3.1) >>
endobj
228 0 obj
(Meta-Learning Scenario)
endobj
229 0 obj
<< /S /GoTo /D (subsection.4.3.2) >>
endobj
232 0 obj
(Experimental Results)
endobj
233 0 obj
<< /S /GoTo /D (subsection.4.3.3) >>
endobj
236 0 obj
(Interpretation of the Learned Functions)
endobj
237 0 obj
<< /S /GoTo /D (section.4.4) >>
endobj
240 0 obj
(Related Work)
endobj
241 0 obj
<< /S /GoTo /D (section.4.5) >>
endobj
244 0 obj
(Summary)
endobj
245 0 obj
<< /S /GoTo /D (chapter.5) >>
endobj
248 0 obj
(A Comparison of Search Algorithms for Heuristic Rule Learning)
endobj
249 0 obj
<< /S /GoTo /D (section.5.1) >>
endobj
252 0 obj
(Implementation of the Algorithm)
endobj
253 0 obj
<< /S /GoTo /D (section.5.2) >>
endobj
256 0 obj
(Search Strategies)
endobj
257 0 obj
<< /S /GoTo /D (subsection.5.2.1) >>
endobj
260 0 obj
(Hill-Climbing)
endobj
261 0 obj
<< /S /GoTo /D (subsection.5.2.2) >>
endobj
264 0 obj
(Beam Search)
endobj
265 0 obj
<< /S /GoTo /D (subsection.5.2.3) >>
endobj
268 0 obj
(Exhaustive Search)
endobj
269 0 obj
<< /S /GoTo /D (section.5.3) >>
endobj
272 0 obj
(Experimental Setup)
endobj
273 0 obj
<< /S /GoTo /D (section.5.4) >>
endobj
276 0 obj
(Results)
endobj
277 0 obj
<< /S /GoTo /D (subsection.5.4.1) >>
endobj
280 0 obj
(Varying the Beam Size)
endobj
281 0 obj
<< /S /GoTo /D (subsection.5.4.2) >>
endobj
284 0 obj
(Single Rules)
endobj
285 0 obj
<< /S /GoTo /D (subsection.5.4.3) >>
endobj
288 0 obj
(Results for Individual Datasets)
endobj
289 0 obj
<< /S /GoTo /D (subsection.5.4.4) >>
endobj
292 0 obj
(Runtime of the Search Methods)
endobj
293 0 obj
<< /S /GoTo /D (section.5.5) >>
endobj
296 0 obj
(Bidirectional Rule Learning)
endobj
297 0 obj
<< /S /GoTo /D (subsection.5.5.1) >>
endobj
300 0 obj
(Theoretical Considerations)
endobj
301 0 obj
<< /S /GoTo /D (subsection.5.5.2) >>
endobj
304 0 obj
(Experimental Setup)
endobj
305 0 obj
<< /S /GoTo /D (subsection.5.5.3) >>
endobj
308 0 obj
(Results)
endobj
309 0 obj
<< /S /GoTo /D (subsection.5.5.4) >>
endobj
312 0 obj
(Discussion)
endobj
313 0 obj
<< /S /GoTo /D (section.5.6) >>
endobj
316 0 obj
(Related work)
endobj
317 0 obj
<< /S /GoTo /D (section.5.7) >>
endobj
320 0 obj
(Summary)
endobj
321 0 obj
<< /S /GoTo /D (chapter.6) >>
endobj
324 0 obj
(A Metric-Based Approach to Regression Rule Learning)
endobj
325 0 obj
<< /S /GoTo /D (section.6.1) >>
endobj
328 0 obj
(Separate-and-Conquer Rule Learning for Regression)
endobj
329 0 obj
<< /S /GoTo /D (section.6.2) >>
endobj
332 0 obj
(Regression Datasets, Regression Algorithms, and Experimental Setup)
endobj
333 0 obj
<< /S /GoTo /D (section.6.3) >>
endobj
336 0 obj
(A Direct Adaption of the SimpleSeCo Rule Learner to Regression)
endobj
337 0 obj
<< /S /GoTo /D (subsection.6.3.1) >>
endobj
340 0 obj
(Splitpoint Processing)
endobj
341 0 obj
<< /S /GoTo /D (section.6.4) >>
endobj
344 0 obj
(Optimizing Several Parameters)
endobj
345 0 obj
<< /S /GoTo /D (subsection.6.4.1) >>
endobj
348 0 obj
(Splitpoint and Left-Out-Parameter)
endobj
349 0 obj
<< /S /GoTo /D (subsection.6.4.2) >>
endobj
352 0 obj
(Parameter of the Regression Heuristic)
endobj
353 0 obj
<< /S /GoTo /D (section.6.5) >>
endobj
356 0 obj
(Results)
endobj
357 0 obj
<< /S /GoTo /D (subsection.6.5.1) >>
endobj
360 0 obj
(Using Different Numbers of Splitpoints)
endobj
361 0 obj
<< /S /GoTo /D (subsection.6.5.2) >>
endobj
364 0 obj
(Comparison with other Systems on the Tuning Datasets)
endobj
365 0 obj
<< /S /GoTo /D (subsection.6.5.3) >>
endobj
368 0 obj
(Comparison with other Algorithms on the Test Sets)
endobj
369 0 obj
<< /S /GoTo /D (section.6.6) >>
endobj
372 0 obj
(Related Work)
endobj
373 0 obj
<< /S /GoTo /D (section.6.7) >>
endobj
376 0 obj
(Summary)
endobj
377 0 obj
<< /S /GoTo /D (chapter.7) >>
endobj
380 0 obj
(Regression via Dynamic Reduction to Classification)
endobj
381 0 obj
<< /S /GoTo /D (section.7.1) >>
endobj
384 0 obj
(Dynamic Reduction to Classification)
endobj
385 0 obj
<< /S /GoTo /D (section.7.2) >>
endobj
388 0 obj
(Experimental Setup)
endobj
389 0 obj
<< /S /GoTo /D (section.7.3) >>
endobj
392 0 obj
(Results)
endobj
393 0 obj
<< /S /GoTo /D (section.7.4) >>
endobj
396 0 obj
(Comparison of Metric-Based Algorithms and Dynamic Reduction)
endobj
397 0 obj
<< /S /GoTo /D (section.7.5) >>
endobj
400 0 obj
(Related Work)
endobj
401 0 obj
<< /S /GoTo /D (section.7.6) >>
endobj
404 0 obj
(Summary)
endobj
405 0 obj
<< /S /GoTo /D (chapter.8) >>
endobj
408 0 obj
(Experiments on Real-World Data)
endobj
409 0 obj
<< /S /GoTo /D (section.8.1) >>
endobj
412 0 obj
(Using Rule Learning Algorithms to Predict Skin Cancer)
endobj
413 0 obj
<< /S /GoTo /D (subsection.8.1.1) >>
endobj
416 0 obj
(Introduction to the domain)
endobj
417 0 obj
<< /S /GoTo /D (subsection.8.1.2) >>
endobj
420 0 obj
(The datasets)
endobj
421 0 obj
<< /S /GoTo /D (subsection.8.1.3) >>
endobj
424 0 obj
(Results)
endobj
425 0 obj
<< /S /GoTo /D (section.8.2) >>
endobj
428 0 obj
(Using Rule Learning to Identify Students Who Need Assistance)
endobj
429 0 obj
<< /S /GoTo /D (subsection.8.2.1) >>
endobj
432 0 obj
(Introduction to the domain)
endobj
433 0 obj
<< /S /GoTo /D (subsection.8.2.2) >>
endobj
436 0 obj
(The dataset)
endobj
437 0 obj
<< /S /GoTo /D (subsection.8.2.3) >>
endobj
440 0 obj
(Results)
endobj
441 0 obj
<< /S /GoTo /D (section.8.3) >>
endobj
444 0 obj
(Related Work)
endobj
445 0 obj
<< /S /GoTo /D (section.8.4) >>
endobj
448 0 obj
(Summary)
endobj
449 0 obj
<< /S /GoTo /D (chapter.9) >>
endobj
452 0 obj
(Discussion of the Results)
endobj
453 0 obj
<< /S /GoTo /D (chapter.10) >>
endobj
456 0 obj
(Conclusions and Future Work)
endobj
457 0 obj
<< /S /GoTo /D (section.10.1) >>
endobj
460 0 obj
(Conclusions)
endobj
461 0 obj
<< /S /GoTo /D (section.10.2) >>
endobj
464 0 obj
(Future Work)
endobj
465 0 obj
<< /S /GoTo /D (subsection.10.2.1) >>
endobj
468 0 obj
(SeCo-Framework)
endobj
469 0 obj
<< /S /GoTo /D (subsection.10.2.2) >>
endobj
472 0 obj
(Classification Heuristics)
endobj
473 0 obj
<< /S /GoTo /D (subsection.10.2.3) >>
endobj
476 0 obj
(Search Algorithms)
endobj
477 0 obj
<< /S /GoTo /D (subsection.10.2.4) >>
endobj
480 0 obj
(Regression Rule Learning)
endobj
481 0 obj
<< /S /GoTo /D (subsection.10.2.5) >>
endobj
484 0 obj
(Real-World Applications)
endobj
485 0 obj
<< /S /GoTo /D (section*.56) >>
endobj
488 0 obj
(Bibliography)
endobj
489 0 obj
<< /S /GoTo /D (section*.59) >>
endobj
492 0 obj
(Own Publications)
endobj
493 0 obj
<< /S /GoTo /D [494 0 R /Fit] >>
endobj
499 0 obj <<
/Length 1029
/Filter /FlateDecode
>>
stream
xڝVr7+p<0 Qђ."3X>PIMQTק
7vB`^_nP9j$uDZy)Ë۫F ZsJ 'ݔFM`H_z64(]mBHR?slUz݇kvξOz[[ŝ!_yĻkxG^j@%]YWY>v\ZO\]@r@jU-E}k(WE/\9_cb*V
Ľq>J
-&dFTfR.Rt}`>Ot'
'?:*wрEEGfIMIXI2fhm6lQMf
_"[pi;upH:qlD?c|K]RZKLp&&}NBjnu2+4⃝gaR|c=cKdvАۍNqޱZP*B۳rT VDL!/MV3@%0N"+R@`JΘ'}n>qx,3#=Hm&p+`L+TGśe !C&-mM15
.
߅@o
3GSGkni4ONYPi;h.)YĆ<~y`<散oS9XtzLJ0i> endobj
496 0 obj <<
/Type /XObject
/Subtype /Form
/BBox [0 0 135.79 63]
/FormType 1
/Matrix [1 0 0 1 0 0]
/Resources 505 0 R
/Length 29
/Filter /FlateDecode
>>
stream
x+2T0 B˥k JG
endstream
endobj
495 0 obj <<
/Type /XObject
/Subtype /Form
/FormType 1
/PTEX.FileName (C:/Program\040Files/MiKTeX\0402.8/tuddesign/tex/latex/tuddesign/logo/tud_logo.pdf)
/PTEX.PageNumber 1
/PTEX.InfoDict 506 0 R
/BBox [0 0 157 63]
/Resources <<
/ProcSet [ /PDF ]
/ExtGState <<
/R8 507 0 R
/R7 508 0 R
>>>>
/Length 29851
/Filter /FlateDecode
>>
stream
xtK8ۤ7U
+54i-!Ԓ)b 끬w/?Fd[
8qy/ץ{_?Ow5hKkg_r^[lۨ]Jjc{_ik.~џnZ>j|)=_Rٗ?:sG߫O}`=c=Vߥ~<?Yq{?1ɸ}{??=
~g04nռgGn|7=u߫_Ͼ9=Oo}awvGz/j_Ypmi߳1wk#}uktOqϞLFNn_v{ozbK.N]/ꞁ/=C=fhTu{vͪ'f{_`}گؿuɹ?0ׯ:T*caiǾ{qO=u/=h~j/Z5}Cޣ|1myzgU͙k%k2'^=z7֛?Q [n,k:ӈy f^cjm/{ MWJ-5F}~ë=h^R/qi%?=g'ƠwoT?O3O]RSh7u~a<{ۋNw3k7߿>iIa!jMn=Tj=.'pO;nON}|aP[}O˗vEO"v{ߋ]v=AmWuk6cu/O^X{Ն{3j[fwcoWWώ7ȩמ:MRn3z32trн'5V_{`F-^*s`c/ FuO-= 5[musr{`1֧u4vsoVR{=bu]@?/Cdz-}kԣU]iKtb{3}{=_kJ;BkiM6ӻy}k@50!,x"GO~>PWx?DG\6Y!E)[:R!Rx]}m;Ė
lR9:dLZ{^]u,ɮмo]{G;%7N{MmcjsNY}zE}-oW!l_"ät+ԔN.rS/PMMb@+f{k |Oڗt_IYtdO,?c<6D<5GbAk{xюt/q==|uaV+
1d~W\dohAҦ+Tc1x46d:6G
散Z{fv6dd>ҽ|cz7>7.3>9m?VT=
OM}iX5Fsvf7աv(5WSaͥۛ]c7н:{e4ڽ>>M6a
MrWOTG=n7)I'>lM陷w<[4^