### Table 6b: Country-Specific Effects of Decentralization on Preference Costs Variable

"... In PAGE 23: ... Therefore, the overall effect of decentralizing maintenance functions is to reduce preference costs, and the optimal degree of decentralization of maintenance activities is 100 percent; where local governments are fully responsible for financing maintenance. Table6 a: Independent Impact of Decentralization on Preference Costs Partial Effects of Functional Decentralization Total Effect of Variable Construction Maintenance Administration Decentralization fiscal -0.15 -0.... In PAGE 24: ...with respect to preference costs. The total effect of decentralization on preference costs is negative but not very significant as can be seen in Table6 a; indicating that decentralization tends to result in lower preference costs. Country specific effects of decentralization on preference costs in Table 6b are large and significant.... In PAGE 24: ... The total effect of decentralization on preference costs is negative but not very significant as can be seen in Table 6a; indicating that decentralization tends to result in lower preference costs. Country specific effects of decentralization on preference costs in Table6 b are large and significant. Most countries have road conditions that are significantly worse than the USA with the exception of Denmark, Germany, UK, Netherlands, Austria, Belgium, Ireland, Italy and Switzerland.... ..."

### Table 1. Classification of global optimization methods based on the degree of history dependence.

"... In PAGE 2: ... Finally, the small energy difference between the correct and incorrect minima and the exponential growth of the density of the non-native states with energy impose strict requirements on the accuracy of energy evaluation (less than about 1 kcal/mol)5. Numerous approaches have been used to attack the global optimization problem in protein structure prediction, with some success1-8 ( Table1 ). These methods are initially classified according to whether they are deterministic or not; stochastic methods are further subdivided according to the degree of similarity between conformations generated in consecutive iterations of the search algorithm.... In PAGE 3: ... Most of the MC-like stochastic global optimization strategies employ a three-step iteration: (i) modify the current conformation by means of a random move; (ii) evaluate its energy; (iii) accept or reject the new conformation according to an acceptance criterion. The random moves can be ranked by magnitude of change with respect to the current conformation ( Table1 ). The first group contains algorithms in which the generated conformations do not depend on the previous ones.... ..."

### Table 5 Algorithmic description of decentralized optimization algorithm.

2001

"... In PAGE 29: ...ontrol outputs and chooses those that minimize the error (e.g. sum of squared difference of the resulting temperature from the desired profile). Table5 and Figure 22 summarize the algorithm and its data flow. Remember that the optimization processes are decentralized, so that each heat source adjusts itself independently, taking a step towards what it thinks minimizes error.... ..."

Cited by 10

### Table 1: Results for Local Search

2005

"... In PAGE 13: ... The local search algorithm starts at a random binary vector and reaches a local maximum in the binary neighborhood by successively moving to the flrst improving neighbor found. Table1 presents the results that were obtained, where average and the maximum objective function value obtained starting from ten random binary vectors are shown for formulations (2) and (6). Matlabr function fmincon uses a sequential quadratic programming ap- proach for solving medium-scale constrained optimization problems.... ..."

### Table 3: Average cut sizes and CPU times (Pentium II - 300 MHz) per local search. SWAP FLIP KL

2000

"... In PAGE 18: ... (1989) is applied to turn the solution into a feasible one. Table3 shows the typical performance in terms of average cut sizes (of 10000 runs) and time per local search (LS) on random starting solutions of the three local search algorithms for four different graphs. The problem instances, computing platforms, and experimental conditions are described in detail in Section 6.... ..."

Cited by 36

### Table 1: The number of optimal solutions, local search cost, and log10-log10 correlation (r) between the number of optimal solutions and local search cost for general JSPs. X Y denotes a mean of X with a std. dev. of Y .

"... In PAGE 8: ...oth SAT and more general CSPs. Singer et al. further demonstrated that the strength of this influence depends critically on the backbone size: the influence is strong in problems with small backbones, but weak in problems with large backbones. In Table1 , we show summary statistics for the number of optimal solutions and the local search cost for our general JSPs. We see both a dramatic drop in the number of solutions and a gradual increase in local search cost as the backbone size is increased.... In PAGE 8: ... This observation is consistent with the increase in search space size for the larger problems, but inconsistent with the observation that square problems are generally much more difficult than rectangular problems. The bottom third of Table1 shows the log10-log10 correlation between the number of optimal solutions and the local search cost. The computed r-values demonstrate that both the number of optimal solutions and the backbone size influence local search cost in the general JSP.... In PAGE 8: ... Cor- relation is relatively strong for problems with small backbones, and drops rapidly as backbone size increases. Although more factors are required to fully explain the variance in local search cost for large-backboned general JSPs, the results in Table1 demonstrate that the interaction effect between backbone size and the number of solutions is not unique to SAT. 5.... In PAGE 8: ... 5.2 Distribution of Backbone Sizes While rectangular JSPs tend to be much easier than square JSPs, this difference was not observed in the local search costs reported in Table1 . In a straightforward experiment, we gen- erated 100 6 4 and 6 6 general JSPs and computed the local search cost for each group, leaving the backbone size uncontrolled.... ..."

### Table 2. Global and local search components used in existing global optimization methods. Method Global Component Local Component

1997

"... In PAGE 9: ... Whenever it is possible, we analyze the balance that each algorithm strikes between global search and local re nement, and relate this balance to its performance. Table2 summarizes this balance for a number of popular global... In PAGE 13: ... This compromise between global and local searches severely limits the performance of global search, and results in very high computational complexity. In view of the drawbacks in trajectory methods and the imbalance between global and local searches ( Table2 ), we propose in Section 4 a new algorithm that uses decoupled global and local search strategies. Our global search strategy is based on a traveling trace that collects geometrical information and uncovers new regions of local minima.... ..."

Cited by 18

### Table 4: Optimal search results for purchasing tasks, n

2007

"... In PAGE 13: ... The first item was the least expensive and is sold by a web site that does not have a P3P policy (thus no privacy information is readily available). Subsequent results increased in order from low to high privacy as the prices increased, as shown in Table4 . Based on previous pilot studies, we found that participants were unlikely to browse beyond the first four search results.... ..."

Cited by 1

### Table 1: Experimental Results for the Local Search Algorithm.

2004

"... In PAGE 12: ... We discuss the impact of this parameter in the next section. Table1 depicts the experimental results on the standard OR Library benchmarks for un- capacitated warehouse location, as well as the M* instances from [21].1 Recall that the M* instances, which capture classes of real UWLPs [21], are very challenging for mathemat- ical programming approaches because they have a large number of suboptimal solutions.... In PAGE 12: ...o note that the algorithm has no prior knowledge of the optimal solution, i.e., it cannot terminate early when the optimum solution is found. As can be seen from Table1 , the algorithm is very robust. It finds optimal solutions with very high frequencies on all benchmarks.... ..."

Cited by 17

### Table 25 identifies shows how changes in each of these dimensions contribute toward achieving the objectives of administrative decentralization.

2004

"... In PAGE 4: ...able 24: New Institutional Arrangements for Improving Access to Justice ............................................... 43 Table25 : Devolving the Employer Function to Provide Administrative Autonomy .... In PAGE 69: ... Table25 : Devolving the Employer Function to Provide Administrative Autonomy e Paying staff from its own budget e Determine the wage envelope Authority to dismiss surplus staff e Control overall staffing numbers Control staffing numbers in local e Establishment Control e offices and facilities Recruitment e Formal employer e Authority to hire e Independent merit-based recruitment mechanism (for example, PSC) Career Management e Promotion e Transfers within local govemment e Horizontal mobility Performance Management e Direct and supervise activities and tasks e Conduct evaluations e Financial rewards/promotion e Ability to disciplinelfire e Set overall wage rates Pay Policy e Set local hardship/remoteness Imnact on Administrative Autonomv J J J J J J J J J J J J J 4 J J J J J J 4 J The Necessitv of Compromises Under a system of complete administrative devolution, all authority for personnel management would be found at the same level of government where staff are located. In practice, there are very sensible reasons to deviate from this simpIistic model.... ..."