Neural networks and the bias/variance dilemma
Neural Computation
Proceedings of the 30th conference on Winter simulation
Computing confidence intervals for stochastic simulation using neural network metamodels
Computers and Industrial Engineering - Special issue on computational intelligence for industrial engineering
Neural Networks for Statistical Modeling
Neural Networks for Statistical Modeling
Simulation Modeling and Analysis
Simulation Modeling and Analysis
Second-Order Learning Algorithm with Squared Penalty Term
Neural Computation
State-of-the-Art Review: A User's Guide to the Brave New World of Designing Simulation Experiments
INFORMS Journal on Computing
Regression models and experimental designs: a tutorial for simulation analysts
Proceedings of the 39th conference on Winter simulation: 40 years! The best is yet to come
A simulation based system for analysis and design of production control systems
Proceedings of the 40th Conference on Winter Simulation
Training feedforward networks with the Marquardt algorithm
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
The construction of a neural network simulation metamodel requires the generation of training data; design points (inputs) and the estimate of the corresponding output generated by the simulation model. A common methodology is to focus some simulation effort in obtaining accurate estimates of the expected output values by executing several simulation replications at each point and taking the average as the estimate. However, with a limited amount of simulation effort available and a rather large input space, this approach may not produce the best expected value approximations. An alternate approach is to distribute that same simulation effort over a larger sample of input points, even if it means the resulting estimates of the expected outputs at each point will be less accurate. We will show through several examples that this approach may result in better neural network metamodels; this conclusion differs from other studies involving regression metamodels.