J. Bresina, M. Drummond, and K. Swanson. Expected solution quality. In Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, pages 1583--1590. Morgan Kaufmann, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....intensive approaches which require significant effort for re tooling to new applications and which include a range of support capabilities beyond simply producing a schedule. Another method is to compare alternative approaches against each other on the same set of scheduling problems (e.g. [1,6,5]) This expedites determining which approach is best, while acknowledging that some support capabilities, such as user interfaces and development environments, will not be evaluated. We conducted a comparative study of several knowledge poor search techniques for scheduling a real world ....

....the system applies all appropriate schedules and calculates all end states. From each of those end states, it generates the next level of possible groups to execute. This process results in all feasible schedules which can be searched by a variety of algorithms to optimize some objective function [1]. Zweben [12] uses simulated annealing to perform iterative repair of schedules in the GERRY system. Constraint based schedulers reduce their search spaces through their hard constraints: at any given point in time, only a subset of possible schedules will meet the hard constraints. Consequently, ....

[Article contains additional citation context not shown here]

John Bresina, Mark Drummond, and Keith Swanson. Expected solution quality. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Montreal, Canada, 1995.

....of time even when the best algorithm for the class of the instance is known. Related work (Bailleux Chabrier 1996) estimate the number of solutions of constraint satisfaction problem instances by iterative sampling of the search tree. In the context of a telescope scheduling application, (Bresina, Drummond, Swanson 1995) use Knuth s sampling method to estimate the number of solutions which satisfy all hard constraints. But the main use of the sampling method is for statistically characterizing scheduling problems and the performance of schedulers. A quality density function provides a background against which ....

Bresina, J.; Drummond, M.; and Swanson, K. 1995. Expected Solution Quality. In Proc. IJCAI-95, 1583-- 1590.

....detrimental effect on the other metric. We have also observed the same trade off. Multi objective problems such as this are often transformed into singleobjective problems by forming a linear combination of the individual objectives. The single objective function used, derived from Bresina (1995) [3], is 11 given as follows: obj = M Gamma M ) oe M (I Gamma I ) oe I (1) where I represents running average inventory, M represents order mean time at dock, while and oe represent the respective means and standard deviations over a set of solutions. The solution set and the objective ....

John Bresina, Mark Drummond, and Keith Swanson. Expected solution quality. In Proceedings of the 14th International Joint Conference on Artificial Intelligence, 1995.

....between the two simulators. The simulators compute two metrics for use in the objective function: mean time at dock and averageinventory. The objective function normalizes these metrics by their means and standard deviations over a set of solutions. The formula, developed using methods reported in [1], is as follows: obj = ai Gamma ai) oe ai (mt Gamma mt ) oe mt where ai represents average inventory and mt represents mean time at dock. Another commonly used performance metric is the makespan. However, the problem with minimizing the makespan is that the production schedule imposes a ....

John Bresina, Mark Drummond, and Keith Swanson, "Expected solution quality," in Proceedings of the 14th International Joint Conference on Artificial Intelligence, 1995.

....total cost obtain from the various schedulers with the costs obtained from iterative sampling. The figure super imposes a histogram from the iterative sampling along with lines that indicate the schedule costs associated with the other algorithms. This follows the expected solution quality [3] method that suggests comparing schedulers with iterative sampling when the cost of an optimal schedule is not known. The graphs make it clear that the good methods produce schedules that are much better than random. In fact, the good methods are so much better than random that the expected ....

John Bresina, Mark Drummond, and Keith Swanson. Expected solution quality. In Proceedings of the International Joint Conference on Artificial Intelligence, 1995.

....are shown in Figure 2. If we express performance improvement in terms of the difference between the objective function scores, then it would be difficult to interpret the significance of the improvement. To overcome this problem, we employ the expected solution quality (esq) methodology ( Bresina,et al. 1995 ] For each JD, 200 randomly generated schedules were scored with the objective function to estimate the quality density function (qdf) Figure 3 plots the standard deviations of the sixty qdfs. The performance improvement for a JD was then expressed in terms of the standard deviation of that ....

John Bresina, Mark Drummond, and Keith Swanson. Expected Solution Quality. Proceedings of IJCAI-95, Montreal, Quebec, 1995.

....analyses are shown in Figure 2. If we express performance improvement in terms of the difference between the objective function scores, then it would be difficult to interpret the significance of the improvement. To overcome this problem, we employ the expected solution quality (esq) methodology [ Bresina, et al. 1995 ] and express the improvement on a JD in terms of the standard deviation of that JD s quality density function (qdf) A qdf is a statistical estimate of the expected density of schedules within different quality ranges and is based on the objective function scores obtained via iterative sampling. ....

John L. Bresina, Mark Drummond, and Keith Swanson. Expected Solution Quality. Proceedings of IJCAI-95, Montreal, Quebec.

....and describe the current operational version of the apa system, focusing on the issues expressed above. Observation Scheduling Domain In this section, we briefly describe the observation scheduling domain. For further details about the apa architecture, see Bresina, et al. 1994) Drummond, et al. 1995), and Edgington, et al. 1996) for a characterization of the scheduling search space, see Bresina, et al. 1995) Our problem domain involves the management and scheduling of ground based, remotely located, fully automatic telescopes. With fully automatic telescopes, the astronomer does not have ....

....Scheduling Domain In this section, we briefly describe the observation scheduling domain. For further details about the apa architecture, see Bresina, et al. 1994) Drummond, et al. 1995) and Edgington, et al. 1996) for a characterization of the scheduling search space, see Bresina, et al. 1995). Our problem domain involves the management and scheduling of ground based, remotely located, fully automatic telescopes. With fully automatic telescopes, the astronomer does not have to be at the observatory and, furthermore, does not have to engage in teleoperation. Fully automatic telescopes ....

[Article contains additional citation context not shown here]

Bresina, J., Drummond, M., & Swanson, K. 1995. Expected Solution Quality. In Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, Montreal, Canada. Morgan Kaufmann Publishers. (Available at url http://ic-www.arc.nasa.gov/ic/projects/xfr/ papers/esq.html.)

No context found.

J. Bresina, M. Drummond, and K. Swanson. Expected solution quality. In Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, pages 1583--1590. Morgan Kaufmann, 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC