Chinese Optics, Volume. 16, Issue 6, 1343(2023)
Multimodal feature fusion based on heterogeneous optical neural networks
Fig. 1. Schematic diagram of the structure of the heterogeneous photonic neural network
Fig. 3. (a) AbsSquared nonlinear activation function structure and (b) the test results
Fig. 7. Schematic diagram of heterogeneous photonic neural network structure based on attention mechanism
Fig. 8. Learning rate and optimizer selection for heterogeneous photonic neural networks based on attentionmechanism
Fig. 9. The effect of random Gaussian noise on the accuracy of the training set
Table 1. Time share of each part of training for heterogeneous electronic neural networks with splicing and fusion
View tableView in ArticleTable 1. Time share of each part of training for heterogeneous electronic neural networks with splicing and fusion
正向传播 反向传播 参数更新时间 总时间 时间/s 6.39 7.58 1.19 15.16 占比/% 42.14 50.00 7.86 100.00
Table 2. Time share of each part of training of heterogeneous electronic neural networks based on the fusion of attention mechanisms
View tableView in ArticleTable 2. Time share of each part of training of heterogeneous electronic neural networks based on the fusion of attention mechanisms
正向传播 反向传播 参数更新时间 总时间 时间/s 6.53 8.18 1.09 15.80 占比/% 41.33 51.75 6.92 100.00
Table 3. Comparison of classification results of advanced methods
View tableView in ArticleTable 3. Comparison of classification results of advanced methods
文献 准确率（%） 文献 准确率（%） 文献[ 25 ]97.18 文献[ 29 ]97.37 文献[ 26 ]92.51 文献[ 30 ]98.10 文献[ 27 ]96.10 文献[ 31 ]98.28 文献[ 28 ]96.00 文献[ 32 ]98.75 基于简单拼接 融合的方法 95.75 基于简单拼接 融合的方法 95.75 基于注意力机制 融合的方法 98.31 基于注意力机制 融合的方法 98.31
Get Citation
Copy Citation Text
Yi-zhen ZHENG, Jian DAI, Tian ZHANG, Kun XU. Multimodal feature fusion based on heterogeneous optical neural networks[J]. Chinese Optics, 2023, 16(6): 1343
Category: Original Article
Received: Mar. 1, 2023
Accepted: --
Published Online: Nov. 29, 2023
The Author Email:
