### Table 11.1: Fixed parameters of the adaptive genetic algorithm program- ming.

### Table 1: Adaptive In-line Genetic Communication Network Programming Algorithm.

"... In PAGE 6: ... Receipt of a fitness function will cause evolution to proceed. From the algorithm description in Table1 , Step 5 is similar to known genetic programming techniques. The primary difference from known genetic programming techniques is that it is occurring in real-time, during normal operation, within a communication network.... ..."

### Table 1: Parameters of the genetic algorithm.

"... In PAGE 4: ... It is with this goal in mind that we chose a simple Braitenberg vehicle, aiming to investigate how the environment will adapt to cope with it; the control program does not change at all. The adaptive algorithm employed is a simple genetic algorithm [7], the parameters of which are given in Table1 (we used the popular freeware software by Grefenstette [8]). (Readers... ..."

### Table 2: Subset Interconnection Design problem using Genetic Algorithms with uniform crossover, crossover rate 0.6, and adaptive mutation. A \* quot; indicates the best. [19] Darrell Whitley, T. Starkweather, and D. Fuquay. Scheduling problems and traveling salesman: The genetic edge recombination operator. In Scha er [17].

"... In PAGE 3: ... No- tice that the GA using adaptive mutation consistently performed better than xed mutation regardless of the crossover method used. Table2 shows the results for a total of 34 runs on eight di erent sets of vertices. The cardinalities of the eight sets of vertices range from n=5 to n=40 in increments of 5.... ..."

### Table 1. Comparison of the two variations of the adaptive simulated annealer and the current best available metaheuristic for the problem, a genetic algorithm for different length runs. % B is the average percent deviation from the best known solutions with 99% con dence intervals shown.

"... In PAGE 11: ... For long searches, there is no difference between the performance of the completely problem independent SA and the SA that uses a problem dependent heuristic to initialize the search. Table1 summarizes a comparison between the two variations of our adaptive sim- ulated annealer and an existing genetic algorithm (GA) for the problem. The speci c GA that we compare to is the current best performing metaheuristic for the problem, which uses a permutation crossover operator known as Non-Wrapping Order Crossover (NWOX) and an Insertion mutation operator [27].... In PAGE 11: ... Approximately 500 generations of the GA take approximately the same length of time to execute compared to a run of SA with 105 total evaluations. For the shortest length run (the rst two rows of Table1 ), the SA with heuristic initial solution nds solutions signi cantly better than when random initial solutions are used (t = 11:1 for a T-test with 1199 dof, p lt; 2 10 27). For the next length run (approximately 0.... ..."

### Table 1 : Option settings step 1 in gatool (genetic algorithm tool) Population Size: 20 Fitness Scaling (scaling function): Rank Selection: Roulette

"... In PAGE 4: ... The reinforcement learning approach was discussed individually by Tom Mitchell [10] and Goldberg [4] in classifier systems. Table1 shows the settings of the options in the genetic algorithm tool of MatLab. The results in figure 4 that shows best and mean fitness =12.... In PAGE 4: ... The results are encouraging compared to Wong [15], Zhan [17], and Jan-Jaap [5]. Finally using reinforcement learning technique, it was observed that the number of children was uniform in evaluation when stochastic selection and adaptive feasible mutation parameters were set in genetic algorithms tool of MatLab package (see Table1 , Figure 4 and Table 2, Figure 5). It was observed that the computations were very expensive as the bidding rate parameter increases from 10 to 20 in reinforcement learning.... ..."

### Table 2: Option settings step 2 in gatool (genetic algorithm tool) Population Size: 20 Fitness Scaling (scaling function): proportional Selection: stochastic uniform Reproduction:

"... In PAGE 4: ... The results are encouraging compared to Wong [15], Zhan [17], and Jan-Jaap [5]. Finally using reinforcement learning technique, it was observed that the number of children was uniform in evaluation when stochastic selection and adaptive feasible mutation parameters were set in genetic algorithms tool of MatLab package (see Table 1, Figure 4 and Table2 , Figure 5). It was observed that the computations were very expensive as the bidding rate parameter increases from 10 to 20 in reinforcement learning.... ..."

### Table 2 genetic, or neural network algorithms. Perhaps more importantly, Fuzzy ARTMAP can be used in an important class of applications where many other adaptive pattern recognition algorithms cannot perform well (see Table 2). These are the applications where very large nonstationary databases need to be rapidly organized into stable variable-compression categories under real-time autonomous learning conditions.

### Table 1. SR, AES, MBF, avg. pm and avg. pc for the TGA at the end of all runs. The self- adaptive pm and pc are not used by the TGA, but are added to the table to see if they are truly random.

"... In PAGE 6: ...5. Table1 indicates that they are indeed random and therefore have no influence on, or are influenced by, the TGA. Genetic Algorithm with Self-Adaptive Mutation (SAMGA) f1 f2 f3 f4 f5 SR 30 0 30 14 0 AES st.... ..."

### Table 2. Genetic algorithm parameters

"... In PAGE 6: ... Figure 5. Discrete workspace of the considered 16-link manipulator The genetic algorithm is implemented using the parameters shown in Table2 in the software Matlab using the genetic algorithm toolbox. The desired end-effector positions are chosen within the region of the manipulator workspace shown in Fig.... ..."