Fig. 5: The multi-task reservoir-computing framework. | Nature Communications

Fig. 5: The multi-task reservoir-computing framework.

From: Wearable in-sensor reservoir computing using optoelectronic polymers with through-space charge-transport characteristics for multi-task learning

Fig. 5

a Illustration of optical multi-task recognition. The optical garment images with associated size information are inputs to the p-NDI optoelectronic reservoir that are classified by memristive organic diode-based readout maps. b The p-NDI circuits performing as reservoir. c The circuit of ionic diodes crossbar performing the readout layer. d Optical images of garments and size labels, sampled from the representative F-MNIST, E-MNIST, and MNIST datasets. e Software pre-processing of the input images to be fed to the organic optical reservoir. f The p-NDI optical reservoir, which extracts discriminative features for all three sub-tasks without training thanks to the fading memory, the task-wise memristive organic diode readout maps, and the multi-task classification flow. g The multi-task recognition result. The system achieved a performance comparable to the software baseline.

Back to article page