Multilayer Memristive Neural Network Circuit Based on Online Learning for License Plate Detection

All‐In‐One Hardware Devices with Event‐Based Vision Sensor Arrays for Image Sensing, Computing, and Learning

“…The voltage signal first flows into the synapse array of the neuron circuit module when the switch composed of the trans-impedance amplifier and resistors turn on. First, the working principle of the synapse array is as the following: [37]…”

“…The voltage signal first flows into the synapse array of the neuron circuit module when the switch composed of the trans‐impedance amplifier and resistors turn on. First, the working principle of the synapse array is as the following: [ 37 ] $$$$\begin{eqnarray} {V}_{-}-{V}_{+}={R}_{f}\ \ensuremath{\times{}}\ \left[\left({G}_{1a}-{G}_{1b}\right)\ \ensuremath{\times{}}\ {V}_{1}+\ \cdots \ +\ \left({G}_{\mathrm{na}}-{G}_{\mathrm{nb}}\right)\ \ensuremath{\times{}}\ {V}_{n}\right]\nonumber\hspace*{-10pt}\\ \end{eqnarray}$$$$where V − and V + are the respective input voltages of the operational amplifiers (hereafter referred to as OAs); G na, b and V n represent the weight parameter of the learning algorithm and input signal, respectively; R f and OAs consist of the trans‐impedance amplifier (hereafter referred to as TIA), which can transform input current into voltage. According to Ohm's laws and Kirchhoff's law, the above equation can achieve multiplying and accumulating (henceforth referred to as MAC) of synapse for input signal and weight parameter.…”

Towards Flexible and Low-Power Wireless Smart Sensors: Reconfigurable Analog-to-Feature Conversion for Healthcare Applications

“…It will help further cut down on energy usage due to the wireless transmission by sending exclusively the classification results. Furthermore, employing analog memristorbased, reconfigurable NN [43][44][45][46] will enable digitizing strictly the classifier's decision while providing the application-or context-specific adaptation. The current order of HAR classification accuracy may suffice for non-critical applications.…”

“…Nevertheless, the advantage of the developed solution lies in its genericity, meaning that it will suit several application cases and remain relatively low-power. Further reduction of energy spent in our acquisition chain can be achieved by employing analog, reconfigurable NNs [44][45][46] based on memristors to digitize the classification result solely. This would also enable learning algorithms such as backpropagation to be implemented on a chip [43], providing the application or context-specific adaptation in the A2F converter and probable mitigation of design constraints by learning and accommodating nonlinearities introduced by analog circuits.…”

Editorial: Advances in non-linear systems and networks

“…Non-linear systems and networks have broad application prospects in the engineering fields of the Internet of Things, medical care, intelligent systems [1][2][3][4][5], etc. With the development of science and technology, according to the current research Frontier, it is not difficult to find that the research fields of non-linear systems and networks are also expanding, including chaotic systems and circuits [6][7][8][9], non-linear device models [10][11][12], memristors [13][14][15][16], neural networks [17][18][19][20][21], neural circuits [22][23][24], synchronous control [25-27] and application research in related fields [28][29][30][31].…”