Results 1  10
of
12
The max karmed bandit: A new model of exploration applied to search heuristic selection
 In AAAI
, 2005
"... The multiarmed bandit is often used as an analogy for the tradeoff between exploration and exploitation in search problems. The classic problem involves allocating trials to the arms of a multiarmed slot machine to maximize the expected sum of rewards. We pose a new variation of the multiarmed bandi ..."
Abstract

Cited by 25 (3 self)
 Add to MetaCart
The multiarmed bandit is often used as an analogy for the tradeoff between exploration and exploitation in search problems. The classic problem involves allocating trials to the arms of a multiarmed slot machine to maximize the expected sum of rewards. We pose a new variation of the multiarmed bandit—the Max KArmed Bandit—in which trials must be allocated among the arms to maximize the expected best single sample reward of the series of trials. Motivation for the Max KArmed Bandit is the allocation of restarts among a set of multistart stochastic search algorithms. We present an analysis of this Max KArmed Bandit showing under certain assumptions that the optimal strategy allocates trials to the observed best arm at a rate increasing double exponentially relative to the other arms. This motivates an exploration strategy that follows a Boltzmann distribution with an exponentially decaying temperature parameter. We compare this exploration policy to policies that allocate trials to the observed best arm at rates faster (and slower) than double exponentially. The results confirm, for two scheduling domains, that the double exponential increase in the rate of allocations to the observed best heuristic outperforms the other approaches.
Nonwrapping order crossover: An order preserving crossover operator that respects absolute position
 Proceedings of the Genetic and Evolutionary Computation Conference (GECCO’06
"... In this paper, we introduce a new crossover operator for the permutation representation of a GA. This new operator— NonWrapping Order Crossover (NWOX)—is a variation of the wellknown Order Crossover (OX) operator. It strongly preserves relative order, as does the original OX, but also respects the ..."
Abstract

Cited by 9 (2 self)
 Add to MetaCart
In this paper, we introduce a new crossover operator for the permutation representation of a GA. This new operator— NonWrapping Order Crossover (NWOX)—is a variation of the wellknown Order Crossover (OX) operator. It strongly preserves relative order, as does the original OX, but also respects the absolute positions within the parent permutations. This crossover operator is experimentally compared to several other permutation crossover operators on an NPHard problem known as weighted tardiness scheduling with sequencedependent setups. A GA using this NWOX operator finds new best known solutions for several benchmark problem instances and proves to be superior to the previous best performing metaheuristic for the problem.
Interaction between reactive and deliberative tasks for online decisionmaking
"... To get actual autonomous engines or systems, it is necessary to equip them with online decisionmaking mechanisms: computation of decisions that fit the current situation, performed in parallel with real execution. However, such a computation introduces a contradiction between requirements on the q ..."
Abstract

Cited by 4 (1 self)
 Add to MetaCart
To get actual autonomous engines or systems, it is necessary to equip them with online decisionmaking mechanisms: computation of decisions that fit the current situation, performed in parallel with real execution. However, such a computation introduces a contradiction between requirements on the quality of the decision and on the time at which it will be delivered. This contradiction is all the stronger as decisionmaking may require intensive computing whose duration is not very well controlled. The solutions that are proposed in the literature to overcome this contradiction are often unsatisfactory, either from the point of view of the quality requirements, or from the one of the temporal requirements, most of the time from the one of the organisation and validation of the whole decisionmaking mechanism.
Using Online Algorithms to Solve NPHard Problems More Efficiently in Practice
, 2007
"... as representing the official policies of the U.S. Government. ..."
Abstract

Cited by 3 (2 self)
 Add to MetaCart
as representing the official policies of the U.S. Government.
A new ant colony optimization approach for the single machine total weighted tardiness scheduling problem
 International Journal of Operations Research
, 2007
"... AbstractIn this paper the NPhard single machine total weighted tardiness scheduling problem in presence of sequencedependent setup times is faced with a new Ant Colony Optimization (ACO) approach. The proposed ACO algorithm is based on a new global pheromone update mechanism, which makes the phero ..."
Abstract

Cited by 3 (2 self)
 Add to MetaCart
AbstractIn this paper the NPhard single machine total weighted tardiness scheduling problem in presence of sequencedependent setup times is faced with a new Ant Colony Optimization (ACO) approach. The proposed ACO algorithm is based on a new global pheromone update mechanism, which makes the pheromone trails asymptotically range between two bounds arbitrarily fixed and the ACO learning mechanism independent of the values of the objective function of the considered problem. Other features of the algorithm include a diversification mechanism for the solution construction phase based on a local pheromone update rule whose effects are restricted to the single iterations, and a cumulative option for the global pheromone update rule. An experimental campaign, carried out on a benchmark available from the literature, was executed to evaluate the proposed ACO and the effectiveness of its optional features. In particular, the obtained results were compared with the ones of a recently proposed ACO algorithm for the same problem by Liao and Juan (2007). The analysis of the outcomes showed the competitiveness of the new ACO approach, which was able to improve about 72 % of the best known results for the benchmark. Finally, a further investigation on a different benchmark set of instances without setup times showed the robustness of the proposed ACO algorithm. KeywordsAnt colony optimization, Metaheuristics, Scheduling, Total weighted tardiness 1.
On the Design of an Adaptive Simulated Annealing Algorithm
"... Abstract. In this paper, we demonstrate the ease in which an adaptive simulated annealing algorithm can be designed. Specifically, we use the adaptive annealing schedule known as the modified Lam schedule to apply simulated annealing to the weighted tardiness scheduling problem with sequencedepende ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
Abstract. In this paper, we demonstrate the ease in which an adaptive simulated annealing algorithm can be designed. Specifically, we use the adaptive annealing schedule known as the modified Lam schedule to apply simulated annealing to the weighted tardiness scheduling problem with sequencedependent setups. The modified Lam annealing schedule adjusts the temperature to track the theoretical optimal rate of accepted moves. Employing the modified Lam schedule allows us to avoid the often tedious tuning of the annealing schedule; as the algorithm tunes itself for each instance during problem solving. Our results show that an adaptive simulated annealer can be competitive when compared to highly tuned, hand crafted algorithms. Specifically, we compare our results to a stateoftheart genetic algorithm for weighted tardiness scheduling with sequencedependent setups. Our study serves as an illustration of the ease with which a parameterfree simulated annealer can be designed and implemented. 1
The Challenge of SequenceDependent Setups: Proposal for a Scheduling Competition Track on One Machine Sequencing Problems
"... Designing a scheduling competition to attract researchers from the several fields interested in scheduling problems seems a challenging, and highly worthwhile effort. In this paper, we propose a design for one possible track of this proposed scheduling competition. Specifically, we propose a track a ..."
Abstract
 Add to MetaCart
Designing a scheduling competition to attract researchers from the several fields interested in scheduling problems seems a challenging, and highly worthwhile effort. In this paper, we propose a design for one possible track of this proposed scheduling competition. Specifically, we propose a track aimed at one machine sequencing problems. We argue that any such track must include problems with sequencedependent setups. Our proposed single machine sequencing track would additionally include a spectrum of objective functions of increasing optimization difficulty under sequencedependent setups. We also offer a problem instance generator along with a set of benchmark problem instances for one potential competition problem—the weighted tardiness scheduling problem with sequencedependent setups.
A Swarm Intelligence Method Applied to Manufacturing Scheduling
"... Abstract—In this paper we present a multiagent search technique to face the NPhard single machine total weighted tardiness scheduling problem in presence of sequencedependent setup times. The search technique is called Discrete Particle Swarm Optimization (DPSO): differently from previous approac ..."
Abstract
 Add to MetaCart
Abstract—In this paper we present a multiagent search technique to face the NPhard single machine total weighted tardiness scheduling problem in presence of sequencedependent setup times. The search technique is called Discrete Particle Swarm Optimization (DPSO): differently from previous approaches the proposed DPSO uses a discrete model both for particle position and velocity and a coherent sequence metric. We tested the proposed DPSO over a benchmark available online. The results obtained show the competitiveness of our DPSO, which is able to outperform the best known results for the benchmark, and the effectiveness of the DPSO swarm intelligence mechanisms.
Online Selection, Adaptation, and Hybridization of Algorithms (Thesis proposal)
, 2006
"... We propose to develop blackbox techniques for improving the performance of algorithms by adapting them to the sequence of problem instances on which they are run. In particular, we propose to develop effective strategies for solving the following four problems: 1. Combining multiple heuristics onli ..."
Abstract
 Add to MetaCart
We propose to develop blackbox techniques for improving the performance of algorithms by adapting them to the sequence of problem instances on which they are run. In particular, we propose to develop effective strategies for solving the following four problems: 1. Combining multiple heuristics online. Given k (randomized) algorithms, we seek to learn online a policy for interleaving and restarting them in order to solve a sequence of instances as quickly as possible. 2. Online reduction from optimization to decision. Given a sequence of instances of an optimization problem and an oracle for the corresponding decision problem, we seek to find a provably (approximately) optimal solution to each instance while minimizing the CPU time consumed by all oracle calls. 3. Online tuning of branch and bound algorithms. Given a branch and bound algorithm with a number of parameters (e.g., choices of relaxations, constraint propagators), we seek to determine nearoptimal parameter settings online. 4. The max karmed bandit problem. In this variant of the classical karmed bandit problem one seeks to maximize the highest payoff received on any single pull, rather than the total payoff. A strategy for solving the max karmed bandit problem can be used to allocate trials among multistart optimization heuristics. For each problem, we prove the existence of a noregret strategy. We show experimentally that our strategy for combining multiple heuristics online has the potential to improve the performance of stateoftheart SAT solvers and A.I. planners, and that our max karmed bandit strategy effectively allocates trials among multistart heuristics for the resourceconstrained project scheduling problem. This proposal is based on part on joint work with Stephen Smith [36, 37] and Daniel Golovin. 1 1