Time-Delay Encoded Image Recognition in a Network of Resistively Coupled VO₂ on Si Oscillators

Abstract

Oscillatory neural networks based on insulator to metal transition of VO2 switches are implemented for image recognition. The VO2 oscillators are fabricated on silicon in a CMOS compatible process. A fully-connected network of coupled oscillators is investigated using programmable resistors as coupling elements. In this approach, input of the image information and data processing is performed in the time domain. In particular, tuning the coupling resistors allows to control the phaserelation between the oscillators. This is used to memorize and recognize patterns in an analog circuit. The concept is demonstrated experimentally on a three-VO2 oscillator network, network whereas simulations are performed on a larger 9-oscillators circuit.

Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks

“…For calculating the power consumption of the circuit, we refer to Shukla et al (2016) , Corti et al (2020) , that demonstrate operations of the oscillators at the power P = 20 μW with a scaled supply voltage <1 V and f = 3 MHz frequency operation. The total energy for the ONN to process one image with 64 filters at 3 MHz, including the waiting time of 5 oscillating period for the output stabilization, is calculated as…”

“…When the network is initialized in this fashion to a phase relation between the oscillators that is different from the stored, stable phase relations, the system relaxes to the nearest stable phase relation, therefore achieving recognition of a pattern. The success of the input time-delay technique for image recognition is explained in detail in Corti et al (2020) . The settling time to the desired output typically varies between 4 and 5 oscillation cycles, nevertheless, after 5 oscillation cycles the phase information becomes stable and can be read-out.…”

Oscillatory Neural Networks Using VO2 Based Phase Encoded Logic

“…In addition, they show good performance in terms of scalability and interconnection with electronic circuits, without requiring any conversion between electrical variables and other nonelectrical variables as occurs with other nano-oscillators that exploit other physical magnitudes. Numerous experimental results of VO 2 nano-oscillators have been reported as well as some preliminary results of applications (Shukla et al, 2016;Corti et al, 2018Corti et al, , 2020Dutta et al, 2019a,b).…”

“…There are contributions of mathematical analysis with simulations using phase models for neurons (Hoppensteadt and Izhikevich, 1999;Follmann et al, 2015) as well as reporting implementations with different types of oscillators [phase-locked loops and voltagecontrolled oscillators (Hoppensteadt and Izhikevich, 2000), nonvolatile logic based on magnetic tunnel junctions (Calayir and Pileggi, 2013), micro-electro-mechanical systems and a feedback loop with transconductance amplifiers (Kumar and Mohanty, 2017), comparator and a digital circuit in Jackson et al (2018), CMOS ring oscillators (Csaba et al, 2016), STOs (Popescu et al, 2018), or VO 2 (Shukla et al, 2014;Maffezzoni et al, 2015;Corti et al, 2018)]. The implementations based on VO 2 devices exhibit potential for very low energy computation (Corti et al, 2020). In the case of electrical oscillators, synapses are implemented with resistors or memristors that play the role of weights.…”

Research on Information Management Based on Image Recognition and Virtual Reality

“…Mireille analyzed the functions of the management information system from the perspective of quality management, and believed that the quality management modules in the system were all composed of ISO9001 templates [6]. Scholars such as Paul Wheatley studied the information management life cycle of digital protection from the perspective of cost accounting under the OAIS framework, and contributed to the research on the long-term preservation of information [7]. In order to achieve effective management of engineering documents, Dr. Zhang et al combined engineering texts with building information models and developed a web-based engineering text management information system using Revit API technology [8].…”