Fig. 3: Monomodal tasks using the VLS PRC.
From: Very-large-scale mimetic optogenetic synapses for physical reservoir computing
a Schematic of handwritten digit recognition. b Classification accuracies of the handwritten digit recognition task after 200 training epochs, with noise levels of σ = 0.05 to 0.4. c Confusion matrix corresponding to the noise level of σ = 0.05. Color intensity is positively correlated with classification accuracy. d Schematic of speech recognition. Speech digits from 0 to 9 are preprocessed using the Lyon passive cochlear model. e Accuracies of speech recognition. f Confusion matrix with noise of σ = 0.05. g Schematic of biological signal recognition with the four types of ECG waveforms (normal sinus rhythm, coronary artery disease, myocardial infarction, and complete atrioventricular block). h Confusion matrix of the ECG classification task. i MSE curve for time-series forecasting task. j Predicted values (red line) compared with ground truth (black line).
