Figure | Results of MT-RDNN for task adaptation. (IMAGE)
Caption
Figure | Results of MT-RDNN for task adaptation. a Conceptual illustration of task adaptation from handwritten digit to letter classification using transfer learning. b Training loss and validation accuracy curves for the English letter recognition task, obtained by fine-tuning the third hidden layer of the MT-RDNN. c Confusion matrix of the MT-RDNN fine-tuned with the GD-based transfer learning. d Performance comparison of the GA, GD, and TFS algorithms in terms of recognition accuracy and number of reconfigured meta-atoms. e Coding patterns of the three hidden layers fine-tuned with the GD-based transfer learning. f Experimental setup of the MT-RDNN for English letter classification. g Measured results of confusion chart for handwritten letter. h Experimental classification results and the corresponding electric field distributions.
Credit
Tie Jun Cui et al.
Usage Restrictions
Credit must be given to the creator.
License
CC BY