### Table 10: The parameters of the evolved EA for TSP.

2005

Cited by 5

### Table 4: Summary of results for the evolved segmental networks. The table gives, for the best solution of each population, the lowest and the highest frequency at which the segmental network can oscillate, the interneurons active in the oscillator with the sign of their in uence and the number of connections of the oscillator. When fewer than 6 interneurons are active, the number of connections after complete removal of the inactive neurons are given.

"... In PAGE 11: ... The solutions have therefore few weak (low weight) connections which do not play an important role in the creation of oscillations. In the case of solutions which use fewer than 8 neurons, the majority of the connections to and from the inactive neurons have been cut; for these solutions Table4 gives the number of connections before and after the removal of the inactive neurons.5 3.... ..."

### Table 5: Summary of results for the evolved segmental networks. The table gives, for the best solution of each population, the lowest and the highest frequency at which the segmental network can oscillate, the interneurons active in the oscillator with the sign of their in uence and the number of connections of the oscillator. When fewer than 6 interneurons are active, the number of connections after complete removal of the inactive neurons are given.

"... In PAGE 12: ... The solutions have therefore few weak (low weight) connections which do not play an important role in the creation of oscillations. In the case of solutions which use fewer than 8 neurons, the majority of the connections to and from the inactive neurons have been cut; for these solutions Table5 gives the number of connections before and after the removal of the inactive neurons.5 3.... In PAGE 63: ...0 - - - - 3.3 Table5 0: Biological connectivity, segmental oscillator, best of run9 0 500 1000 1500 2000 2500 3000 0 0.2 0.... In PAGE 64: ...9 - - - - 3.3 Table5 1: Biological connectivity, segmental oscillator, best of run10 0 500 1000 1500 2000 2500 3000 0 0.2 0.... In PAGE 66: ...0 [3, 9] - - 5.0 Table5 2: Biological connectivity, inter-segmental coupling, best of run1 0 20 40 60 80 100 120 140 0 0.5 1 1.... In PAGE 67: ...0 [4, 1] - - 5.0 Table5 3: Biological connectivity, inter-segmental coupling, best of run2 0 20 40 60 80 100 120 140 0 0.5 1 1.... In PAGE 68: ...0 [5, 10] - - 5.0 Table5 4: Biological connectivity, inter-segmental coupling, best of run3 0 20 40 60 80 100 120 140 0 0.5 1 1.... In PAGE 69: ...0 [2, 10] - - 5.0 Table5 5: Biological connectivity, inter-segmental coupling, best of run4 0 20 40 60 80 100 120 140 0 0.5 1 1.... In PAGE 70: ...0 [11, 1] - - 5.0 Table5 6: Biological connectivity, inter-segmental coupling, best of run5 0 20 40 60 80 100 120 140 0 0.5 1 1.... In PAGE 72: ...1 5.0 Table5 7: Biological connectivity, sensory feedback, best of run1 EINl CCINl LINl EINr CCINr LINr ECl ECr BS EINl 0.4 - - - -2.... In PAGE 72: ...3 5.0 Table5 8: Biological connectivity, sensory feedback, best of run2 EINl CCINl LINl EINr CCINr LINr ECl ECr BS EINl 0.4 - - - -2.... In PAGE 72: ...5 5.0 Table5 9: Biological connectivity, sensory feedback, best of run3 EINl CCINl LINl EINr CCINr LINr ECl ECr BS EINl 0.4 - - - -2.... ..."

### Table 4: Stack Primitives Essential to All Evolved Solutions

1996

"... In PAGE 18: ... From Table 3 we can identify ve primitives which are essential to the operation of all four evolved solutions and ve pairs of primitives where one or other is required. These are shown in the two halves of Table4 together with the number of runs where they were removed from the population by 21 generation equivalents (i.e.... In PAGE 18: ...emoved from the population by 21 generation equivalents (i.e. by the point where all four solutions had evolved). After the equivalent of 21 generations in 43 of 60 runs, the number of one or more of the tree-primitives shown in the left had side of Table4 had fallen to zero. That is the population no longer contained one or more primitives required to evolve a solution (like the solutions that have been found).... In PAGE 18: ... That is the population no longer contained one or more primitives required to evolve a solution (like the solutions that have been found). In 12 of the remaining 17 populations both of one or more of the pairs of primitives shown on the right hand side of Table4 had been removed from the population. Thus by generation 21 in all but 5 of 60 runs, the population no longer contained primitives required to evolve solutions like those found.... ..."

Cited by 1

### Table 1: The six variants of collect.

"... In PAGE 5: ... This yields three options for defining the operation duration, namely probing that accepts an empty state, single that is satisfied with a non-empty state response, and persistent that remains active until explicitly cancelled by the application. The combinations of the space and time dimensions yield six variations of the collect operation, outlined in Table1 . The seman- tics of publish and subscribe are unaltered.... ..."

### Table 6. Comparing the variants of COS - Removed edges.

"... In PAGE 24: ... 6.1 Minimizing the number of stubs Table6 provides a direct comparison of the Briand and COS approaches to edge selection for breaking cycles in an ORD, and hence to stubbing. There are three aspects to that comparison.... In PAGE 25: ... We conclude that using Briand-A is preferable due to its single pass, and in the remainder of the results described here we utilize Briand-A as the representative of the Briand approach. Comparing the lines in Table6 corresponding to Briand-A and to COS-A with costs (9999, 2, 9999, 2, 2, 9999) allows a direct comparison of the two methods when they are applied only to ADP edges. Those results show no substantial difference between the two methods, with Briand-A being superior in two test cases, COS-A being superior in three test cases and the two methods being equal in one test case.... In PAGE 25: ... Recall that the Briand weights are based on the product of in and out degrees, while the COS weights are based on the edge types. The third aspect of comparison arising from Table6 comes in comparing Briand-A to COS-A with costs (2,2,2,2,2,2). This measures the advantage that can be gained in terms of the total number of stubs that need to be written by allowing the stubbing of all types of edges, rather than just ADP edges.... In PAGE 27: ... This is not surprising since that cost model assigns a weight of 2 to ADP edges, and a weight of 20 to inheritance edges. Even so, for the largest application Drawserv, COS-A selected 15 inheritance edges for removal, and this resulted in removing about 700 fewer ADP edges than if no inheritance edges were removed (from Table6 ). Note also that no composition edges were selected for removal in any test case, since as noted earlier, no composition edges are located in a cycle in any test case.... In PAGE 27: ... 6.3 Which version of COS? The results in Table6 allow a comparison of the effectiveness of the three versions of the COS algorithm. In this table we show the number of edges (individual edges) removed by the COS variants using two cost models.... ..."

Cited by 1

### Table 2. Comparison of results obtained from ACSGA-TSP and ACS-TSP algorithms on the large TSP instance rat783. Results given are averages of best tour lengths over 3 runs, standard deviation values for the results, and the number of iterations. In (a), the algorithms were run such that the same number of tours were explored as in ACS-TSP runs. In (b), the algorithms were run for the same amount of time as in ACS-TSP runs.

2002

"... In PAGE 3: ... Thus, in its current form, ACSGA-TSP would not outperform the ACS-TSP algorithm. The rate at which good solutions were found was observed to be quicker using ACSGA-TSP, as seen in Table2 . Quick convergence to good solutions is thus a desirable characteristic of the ACSGA-TSP algorithm.... In PAGE 3: ... Quick convergence to good solutions is thus a desirable characteristic of the ACSGA-TSP algorithm. For large problems, where optimal solutions are intractable or not desired, this algorithm provides good solutions faster than the ACS-TSP algorithm, as shown in Table2 . This characteristic is mainly due to the variety in the population of ants which facil- itates early exploration of the search space.... ..."

Cited by 12

### Table 2: Relative performance on the 20-city-TSP, 50 problems

"... In PAGE 7: ... As is customary,we used the minimum spanning tree #5B4#5D of the remaining cities to estimate the completion cost of the partial tour. Table2 shows the experimental results with city coordi- nates drawn from the intervals #5B1; 50#5D and #5B1; 100#5D. The per- formance is given relativetoIDA* in terms of node expan- sions and CPU time consumption.... In PAGE 7: ... As can be seen, neither of the node ordering heuristics #28PV+Sort or History#29 yields substantial performance improvements. This is not surprising, since only 8#25 of the total nodes are visited in the last iteration, yielding an upper bound on the maximal improvement that can be achieved byany kind of node or- dering #28see the last line in Table2 #29. The History results are based on a 2-dimensional history table that holds for every city pair the frequency it contributed to the longest tour.... In PAGE 7: ... As is always the case in applications where the cost bound increase is not known a priori, cut o#0Bs are only feasible when the retrieved cost bound is higher than the temporary candidate for the next cost bound. Table2 also shows how further savings are achieved with the more sophisticated hashing techniques. The Trans+ReHash variant resolves storage collisions by giving preference to states encountered in the shallow graph levels near the root.... In PAGE 7: ... Figure 5 illustrates in a more general way the in#0Duence of the tree characteristics on the relative search e#0Eciency. The data shown is that from Table2 , but extended to include information from all four domain intervals con- sidered. Instead of plotting the performance relativeto the domain of the city coordinates, we took the heuristic branching factor achieved with the simple IDA* as a per- formance measure.... In PAGE 8: ... Although the simple successor ordering of Sort did not pay o#0B, the other heuristics, namely PV, Trans and Trans+Move, reduce the search time by 13, 24 and 37#25, respectively. These results compare favorably to those of others #5B25 Table2 #5D, #5B2, 12, 21, 22, 29#5D. In practice, one would #0Crst include the PV-heuristic, be- cause of its negligible space and time overheads.... ..."

### Table 6: List of 2 DF in distinct solutions obtained after deleting undetermined parameters; NExp = 411 experimental points. (To save the space only the best unphysical solutions of the \DRNbior quot; variant are shown.)

"... In PAGE 24: ...ameters had no influence on the above conclusions 1.{4. whereas the number of parameters in e ect had been considerably reduced. The new inferences derived from the Table6 read: 5. The considered data base requires a complicated model for its description.... In PAGE 26: ... The parameter g1 of the paper [21] is found to be relatively important, the D{waves parameters g2, g3 being much less necessary. For illustrations we have chosen the solution from the \DRNbior quot; variant with 2 DF =1:16 (see Table6 ). The data on total cross sections and the theoretical curves in terms of the quasi{amplitude (15) are drawn in Fig.... In PAGE 32: ...quantities D1 and D2 found in the solution with 2 DF =1:16 (see the variant \DRNbior quot; of the Table6 for values of scattering lengths). We also quote here the predictions of papers [12, 57, 13] and the results ALL and TRI of the linear t from the Table 7 for an easy comparison.... ..."

### Table 3. Information repositories evolved by specialists versus those evolved by knowledge workers. Evolved by specialists Evolved by knowledge workers

2001

"... In PAGE 6: ... These breakdowns lead the users to continually and directly evolve and refine their information space, without relying on professionals. Sustaining the usefulness and usability of living information repositories over time involves important challenges and trade-offs (summarized in Table3 ). These trade-offs depend on whether these information repos- itories are evolved by specialists or by knowl- edge workers.... ..."

Cited by 28