### Table 5: Results of the capacitated facility locations problems on the AP1000

### Table 2. Opportunities and problem solved by the e-newspaper

"... In PAGE 8: ...1 Relative advantage In the prototype test, the 36 users were asked what opportunities they envisioned with the e-newspaper and what problems the e-newspaper solves. In Table2 the answers are clustered into four categories, some answers are mapped into more than one category. ... ..."

### Table 3. Results with SA for solving the RND problem (omnidirectional BTSs).

"... In PAGE 6: ... Therefore, we decided to keep the first at a constant value of 50, and allow the tuning of the latter. Table3 shows the results. The tests have been performed in a 16 machine cluster named in dedicated mode, and the code has been developed using the MALLBA library [17].... ..."

### Table 4. Results with CHC for solving the RND problem (omnidirectional BTSs).

"... In PAGE 6: ...4 Results with CHC When using the CHC algorithm on the same instance that the previous methods, we have considered two parameters that can be tuned (the rest of configuration parameters are fixed to their typical values): population size and cataclysmic mutation probability. Table4 shows the best configuration, and the results obtained. The tests have been performed in a 16 machine cluster named in dedicated mode, and the code has been developed using the MALLBA library [17].... ..."

### Table 1: Formal descriptions of the synthesis problems

"... In PAGE 2: ... Figure 1 is a basic structure of a neural network strategy. Figure 1: The architecture of a neural network strat- egy The design principles of a neural network strategy are: the inputs of the neural networks are the matrix of multiple solutions from ESs (refer to Table1 ); the outputs of neural networks should be the nal solu- tions after synthesizing multiple solutions. If a neural network can converge for all of patterns after training, this neural network can act as an inductive function.... ..."

### Table 1: Node counts and time for instances of multi-commodity network ow problems CPLEX CPLEX + CUTS

2007

"... In PAGE 126: ... CPLEX branch-and-bound was used to solve the two mixed integer programming formulations. Table1 1: Comparison of two formulations: lower and upper bounds were returned at the end of 300s of computation time (P1) (P2) Prob LB UB LB UB E10 10 0.00 0.... In PAGE 127: ... This shows that as an integer programming formulation, with no additional cuts or heuristics added, formulation (P 2) performs better than formulation (P 1). Table1 2: Comparison of two formulations: Node counts and solve times (P1) (P2) Prob Node Count Time Node Count Time E10 10 240 1.5 56 0.... In PAGE 128: ... The time limit was 300s, so if optimal solution is not found in the allotted time for a problem the corresponding entry for solve time is 300s and node count entry is the number of nodes explored in 300s. Table1 3: Comparison of two formulations with cutting planes and heuristics: lower and upper Bounds after 300s of computation time (P1) (P2) Prob LB UB LB UB E10 10 0.00 0.... In PAGE 128: ...00 0.00 Entries in bold represent that optimal solution was found in 300 second Looking at the results from Table1 3, we can see that, with the help of cuts and heuristics, formulation (P 1) was able to provide better results than (P 2). More problems were solved to optimality and for except one, the bounds provided for the problems not solved to optimality in allotted time by formulation (P 1) were stronger than formulation (P 2).... In PAGE 129: ...Table1 4: Comparison of two formulations with cutting Planes and heuristics: node counts and computation times (P1) (P2) Prob Node Count Time Node Count Time E10 10 0 4.01 0 0.... ..."

### Table 3.3 Cluster analysis proximity matrix for the #1 problem solving/default condition, all products combined

### Table 4. Clustering

"... In PAGE 8: ... Hard problems solved After the competition a clustering of the solvers was presented, in which solvers are grouped together if they behave similarly on a given set of benchmarks. Table4 shows which solver was most similar to MidiSAT for which benchmark class, i. e.... ..."

### Table 5 shows some additional results summed over families of problems. Most of these families came from the 2002 SAT competition. The families 03 rule and 06 rule come from the IBM Formal Verification Benchmark suite (we did not run Zchaff on these families). The number in brackets after the family name is the number of problems in the family. The time given is the total time to solve all problems in the family in CPU seconds. In these experiments a single problem time-out of 20,000 CPU seconds was imposed, and 20,000 was added to the total time to solve the family for each failed problem. The number in brackets after the total time is the number of problems the solver failed to solve. For the preprocessor failure to solve means that the problem was not resolved at preprocessing time. The first set of times is the time required by the

2003

"... In PAGE 11: ...11 Table5 Performance on various families of problems. Time in CPU seconds.... ..."

Cited by 39