Estimating the Significance of Computer Model Factors Based on a Simple Neural Network
Introduction. The existing means of computer technology provide an opportunity to develop models of any complexity. This circumstance provokes developers of computer models of complex systems to excessive detailing. Among experienced specialists in computer modeling, there is an opinion that sometim...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
V.M. Glushkov Institute of Cybernetics
2024-12-01
|
| Series: | Кібернетика та комп'ютерні технології |
| Subjects: | |
| Online Access: | http://cctech.org.ua/13-vertikalnoe-menyu-en/651-abstract-24-4-3-arte |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Introduction. The existing means of computer technology provide an opportunity to develop models of any complexity. This circumstance provokes developers of computer models of complex systems to excessive detailing. Among experienced specialists in computer modeling, there is an opinion that sometimes at the preliminary stage of model development, the number of insignificant factors can reach 80%. Such an increase in dimensionality not only significantly complicates the implementation of computer experiments, but can also have a significant impact on the understanding of the interaction of important factors that determine the basis and essence of the functioning of a complex system. Therefore, it is no less important for further model research, and especially for the optimization of a complex system, to determine insignificant factors.
The purpose of the work is to develop an algorithm for determining insignificant factors in the presence of a set of training data, in which the number of data samples is relatively small and exceeds the number of factors by only 2-3 times. For this, a neural network model implementing regression created using the Keras library was used. Artificially created datasets were used to conduct experiments to determine network parameters (number of layers, number of hidden neurons in a layer, as well as the number of learning epochs).
The results. The resulting neural network model demonstrated effective performance on test data sets. The model was then used to determine the significance of factors in sets of initial populations for a multipopulational genetic algorithm (MGA) study.
Conclusions. The proposed algorithm based on a simple neural network allows to correctly and quickly determine insignificant factors in a set of initial populations for the study of MGA, containing from 8 to 10 populations (250 - 300 samples). Since the initial weights of the neural network are chosen randomly, the results of different runs on the same set of data are slightly different. Therefore, in the general case of evaluating the significance of factors of a computer model, several runs must be made to obtain more reliable results. |
|---|---|
| ISSN: | 2707-4501 2707-451X |