Convolutional Neural Network Integration to a 3-D Ray-traced Biological Neural Network
Date
2024
Authors
Chen, Xuan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This study presents a machine learning model that integrates a Convolutional Neural Network (CNN) into a 3-D Ray-traced Biological Neural Network (RayBNN). RayBNN specializes in rearranging and adapting to various problems through transfer learning. CNN is renowned for extracting features from data efficiently, which may boost to RayBNN performance. In this report, two integration schemes are implemented and tested on the Modified National Institute of Standards and Technology(MNIST) and Wakefulness Test recordings (MWT) datasets respectively. Using MNIST dataset, we trained a CNN with an artificial neural network (ANN) and an auto encoder/decoder to extract features from datasets and used them as the input of RayBNN. Using the CNN with ANN approach, an accuracy of 0.9919 ± 0.0012, a precision of 0.9921 ± 0.0008, a recall of 0.9922 ± 0.0012 and an F1-score of 0.9920 ± 0.0008 were obtained. When using CNN with auto encoder/decoder for feature extraction, the accuracy, precision, recall, and 0.9905 ± 0.0035, 0.9901 ± 0.0050, 0.9881 ± 0.0068, and F1-score at 0.9908 ± 0.0010 respectively. For MWT dataset, Cohen’s Kappa values of 0.68 ± 0.05, 0.71 ± 0.04, 0.04 ± 0.02, and 0.06 ± 0.02 for Wakefulness, Microsleep Episode, Microsleep Episode Candidate, Episodes of Drowsiness classes were obtained using CNN with ANN to extract features for RayBNN.
Description
Keywords
deep learning, Convolutional Neural Network, Biological Neural Network, transfer learning