Fig. 5: Multimodal tasks using the VLS PRC.
From: Very-large-scale mimetic optogenetic synapses for physical reservoir computing
a Schematic of a speech-to-image task using the 3D VLS PRC. b Loss curves of the speech-to-image task. c–f Left: digital images generated (c) in darkness, (d) under weak illumination, (e) under intense illumination, and (f) after second-time training. The blue and red boxes label the irregular images and errors. Right: visualization of the latent vectors with reduced dimensions. g, h Blood cell images generated (g) in darkness, (h) under intense illumination. The blue and red boxes label the irregular images and errors. i, j Colorectal tissue images (i) in darkness, (j) under intense illumination. The blue and red boxes label the irregular images and errors.
