"... In PAGE 12: ... All experiments were performed on the SPECint95 benchmarks. Table2 lists the benchmarks and the input sets 1. All simulations were run until completion.... ..."

"... In PAGE 12: ... All experiments were performed on the SPECint95 benchmarks. Table2 lists the benchmarks and the input sets 1. All simulations were run until completion.... ..."

"... In PAGE 12: ... Another important contribution from this work is the introduction of a set of 13 exam timetabling problems, which became standard benchmarks in the eld. They have been widely studied and used by di erent approaches during the years (see Table6 ). We call this the University of Toronto data and discuss it further in Section 3.... In PAGE 30: ... This variant is named Toronto e in Table 5. The approaches developed and tested on di erent variants of the Toronto datasets during the years are listed in Table6 (ordered by the year in which the work was published). The values in \() quot; following the variants of the data give the number of problem instances tested by the corresponding approaches.... In PAGE 31: ... Table6 . Approaches to the Toronto Benchmark Datasets Reference Approach/Technique Problem Carter et al Graph heuristics with clique initialisation and a(13), b(13) [54] 1996 backtracking Carter amp;Johnson Almost cliques with su cient density as the a(13) [51] 1996 initialisation for graph heuristics Burke et al Memetic Algorithm with hill climbing and light and c(5) [40] 1996 heavy mutation Burke et al Di erent initialisation strategies in Memetic d(3) [41] 1998 Algorithms measured by diversity Burke et al Non-determinisms introduced by selection strategies d(3) [42] 1998 in graph heuristics Burke amp;Newall Multi-stage Evolutionary Algorithm based on d(3) [37] 1999 Memetic Algorithm Terashima Genetic Algorithm with in-direct coding of the e(12) -Marin et al constructive strategies and heuristics [149] 1999 Caramia et al Iterated algorithm with novel improving factors a(13), b(13), [48] 2001 c(5) Di Gaspero [73] Adaptive tabu list and cost function in Tabu Search b(11), c(5), amp;Schaerf 2001 d(3) Di Gaspero Multiple neighbourhood Tabu Search b(7),d(3) [72] 2002 White amp;Xie [156] Tabu Search with long term memory b(2) [157] 2001 amp;2004 Relaxation on long and short term tabu lists b(7) Paquete amp; [120] Tabu Search with Lex-tie and Lex-seq strategies in b(8) Stutzle 2002 the objective function Merlot et al Constraint programming as initialisation for a(12), b(12), [110] 2003 Simulated Annealing and hill climbing c(5), d(2) Casey amp; [55] GRASP with modi ed Saturation Degree b(10) Thompson 2003 initialisation and Simulated Annealing improvement Burke amp;Newall Great Deluge with adaptive ordering as the b(11) [38] 2003 initialisation Burke amp;Newall Graph heuristics with adaptive heuristic modi er to b(11), d(3) [39] 2004 dynamically order the exams Burke amp;Bykov Time-prede ned Great Deluge and Simulated b(13), d(2) et al [18] 2004 Annealing Asmuni et al Fuzzy rules with Largest Degree, Saturation Degree b(12) [8] 2004 and Largest Enrolment [9] 2006 Fuzzy evaluation function with multiple criteria b(12) Ross et al Genetic Algorithm evolving constructive strategies e(12) [138] 2004 and heuristics Burke et al Hybridising graph heuristics in hyper-heuristic by b(4) [22] 2005 CBR and systematic strategies Cote et al Bio-objective Evolutionary Algorithm with local b(12) [63] 2005 search operators in the recombination process Kendall amp;Hussin Tabu Search based hyper-heuristic b(8) [94] 2005 Yang amp;Petrovic Similarity measure using fuzzy set on selecting b(12) [161] 2005 hybridisations of Great Deluge and graph heuristics Abdullah et al Large neighbourhood search with tree-based b(12), c(5) [3] 2007 neighbourhood structure [4] 2007 Tabu Search based large neighborhood search... ..."

"... In PAGE 4: ... (8) 3 Numerical Evaluation For our experimental evaluation we used a collection of nine publicly available datasets, part of the UCI repository4. A summary of these datasets is shown in Table1 . These datasets are commonly used in machine learning as a benchmark for performance evaluation.... ..."

"... In PAGE 4: ... (8) 3 Numerical Evaluation For our experimental evaluation we used a collection of nine publicly available datasets, part of the UCI repository4.A summary of these datasets is shown in Table1 . These datasets are commonly used in machine learning as a benchmark for performance evaluation.... ..."

"... In PAGE 6: ... We performed our evaluations on the Cora dataset, plus 18 other text datasets provided by Han and Karypis (2000). Refer to Table1 . The classification tasks are drawn from standard benchmarks such as Reuters, OHSUMED, and TREC, among others.... ..."

"... In PAGE 7: ... We performed our evaluations on the Cora dataset, plus 18 other text datasets provided by Han and Karypis (2000). Refer to Table1 . The classification tasks are drawn from standard benchmarks such as Reuters, OHSUMED, and TREC, among others.... ..."

"... In PAGE 6: ... For each benchmark, the table lists the number of ST and MT branches (only jsr are classi ed as ST since all jmp instructions are conservatively considered to be MT). The dynamic characteristics of the benchmarks are shown in Table3 . They include the input le, the size of the trace, the percentage of indirect branches (IB), the percentage of indirect branches that are MT and ST indirect branches, the number of indirect branches executed per instruction and per conditional... In PAGE 7: ... branch, and the average number of targets per indirect branch. As we can see from Table3 , most benchmarks execute more ST branches than MT branches. This is expected since ST branches are library function calls.... ..."