Hybrid MoS2/PbS Quantum Dots Toward In-Sensor Reservoir Computing

Neuromorphic visual systems are potential to overcome the von Neumann bottleneck, emulating the human retina to implement in situ vision preprocessing. Specifically, in-sensor reservoir computing (RC) is capable of processing spatio-temporal vision information with high efficiency. However, traditional photodetectors are incapable of the fading memory ability, which is the core of the RC systems for temporal information processing. Here, we report a hybrid MoS2/PbS quantum dots based optoelectronic device with short-term memory properties for RC demonstration. The photogating effect induced dynamic temporal photoresponse behavior can be effectively regulated by the incident optical intensity and time. As a proof of concept, image classifying with a high accuracy of 95.8% and time series prediction are demonstrated, paving a way for intelligent edge computing.

Automated summary

The researchers report a hybrid MoS2/PbS quantum dots-based optoelectronic device with short-term memory properties for in-sensor reservoir computing (RC). They demonstrate that the device can effectively regulate the dynamic temporal photoresponse behavior through the photogating effect induced by incident optical intensity and time. The device shows high accuracy in image classifying and time series prediction, indicating its potential for intelligent edge computing.