16+
DOI: 10.18413/2518-1092-2023-8-4-0-5

DESIGN OF RECURRENT NEURAL NETWORKS FOR CLASSIFICATION OF AGE DIFFERENCES IN THE FUNCTIONING OF THE SYMBOLIC SYSTEM OF QUANTITY ASSESSMENT

The article presents the result of the development and training of 4 recurrent neural network architectures to solve the problem of classifying age-related differences in the functioning of the symbolic system of quantity assessment. When designing neural networks, some effective algorithms were used: cells with long short-term memory, a modification that allows a signal to be fed to the neural network in forward and reverse order, and preliminary 1D convolutions of the signal before feeding it to recurrent layers. The best result on all data sets was demonstrated by a recurrent neural network with signal pre-convolution layers. Accuracy varies between 86-88% depending on the dataset. The specified accuracy was obtained on data to which the baseline correction algorithm was applied.

Number of views: 363 (view statistics)
Количество скачиваний: 1273
Full text (PDF)To articles list
  • User comments
  • Reference lists
  • Thanks

While nobody left any comments to this publication.
You can be first.

Leave comment: