A quantum dot image processor

Electrochemically self-assembled nanostructure arrays

“…In Ref. [7] and references therein, it was shown that ordered arrays of nanowires that exhibit NDR can lead to an extremely powerful neural network architecture capable of associative memory tasks, Boolean logic, and image processing. Additionally, if the NDR can be modulated by an external agency, such as an infrared source, then that can lead to more complex architectures that mimic, for instance, the Fitzhugh-Naguma model of impulse propagation along nerve membranes [12].…”

“…These structures have so far shown strong optical non-linearity [3] (with applications in optical switching, mixing, frequency conversion, limiting, etc. ), stable electronic bistability for non-volatile memory applications [4], room-temperature infrared photodetection (in the wavelength range 2-5 mm) [5], room-temperature single electron charging (Coulomb blockade and Coulomb staircase) [6], and even negative differential resistance (NDR) in a configuration that can ultimately lead to revolutionary neuromorphic computing architectures and superfast high-density image processors [7]. In this paper, we describe briefly electrochemical nanosynthesis and report on an intriguing self-assembled device-an infrared photodetector that also exhibits an NDR.…”

Non-Boolean Computing with Spintronic Devices

“…Another direction that has been popular is the cellular neural network (CNN) computing model [36,35,136,156,45,62,200,129,141]. Quantumdot (QD) arrays have been explored [118,117,93] for low-level image processing applications. DWAVE's [8, 56] recent work uses quantum annealing to solve optimization problems at ultra-low temperature.…”

An experimental demonstration of the viability of energy minimizing computing using nano-magnets

“…These vision problems place high demand on computational resources (on Boolean logic based computing platforms). There have been proposals for using regular arrays of quantum-dots [17], [18] and nano-magnets [19] for low-level vision, mainly segmentation where the input and the output are both regular grid of pixels. In this work we consider quadratic energy forms that arise in contexts involving extended image features rather than individual pixels.…”