A neuron circuit based on N-type locally active memristor

Abstract

The present invention discloses a neuron circuit based on an N-type local active memristor, which comprises an N-type local active memristor, an inductor L, an excitation voltage source V D The N-type local active memristor (LAM) is a type of memristor that exhibits N-type negative differential conductance in its DC V-I characteristic curve. By connecting this N-type LAM with an inductor in series to form a neuron circuit and applying a suitable voltage stimulus, the N-type LAM is able to amplify small signals and generate complex neuromorphic behavior, laying a foundation for its application in neuromorphic computing.

Description

Technical Field

[0001] The present invention relates to the field of neuron circuit design, and in particular to a neuron circuit composed of an N-type local active memristor. Background Art

[0002] As demands for computing system speed, integration, and energy efficiency continue to rise, traditional von Neumann-based computing architectures are increasingly unable to meet these demands. However, neuromorphic computing, which draws inspiration from the human brain's information processing methods, demonstrates enormous potential due to its high energy efficiency and large computing capacity. The exploration of hardware implementations of neurons, the computing units of neuromorphic computing, is in full swing.

[0003] Professor Cai Shaotang pointed out that complex neuromorphic behaviors exist in local active regions. The nonlinear and local active properties of locally active memristors can be used to design neuronal circuits and apply them to neuromorphic computing.

[0004] Locally active memristors (LAMs) exhibit a negative differential resistance (conductance) region in their DC VI characteristics and can be categorized as either S-type (current-controlled) or N-type (voltage-controlled). Neuron circuit architectures based on memristors such as NbOx, TiO2, and Mott have been widely proposed and experimentally verified. However, all of these memristive devices are S-type LAMs. Neuron circuits based on N-type LAMs have yet to be demonstrated. Summary of the Invention

[0005] The present invention aims to solve the problems existing in the prior art and proposes a neuron circuit model based on an N-type local active memristor to simulate the neuromorphic behavior of neurons.

[0006] The neuron circuit proposed in the present invention comprises an N-type local active memristor, an inductor L, and an excitation voltage source VD; the negative electrode of the excitation voltage source VD is connected to the ground, the positive electrode is connected to one end of the inductor L, the other end of the inductor L is connected to one end of the N-type local active memristor, and the other end of the N-type local active memristor is grounded;

[0007] The mathematical model of the N-type local active memristor is:

[0008]

[0009] Where i, v, and x are the current flowing through the memristor, the voltage across the memristor, and the state variable of the memristor, respectively. d1, d0, τ, and α are all constants. G M is the memetic derivative function.

[0010] Beneficial Effects: This invention proposes a simple neuron circuit consisting of only an N-type locally active memristor, an inductor, and an excitation voltage signal. The N-type locally active memristor exhibits characteristics such as local activity, high integration, and nonlinearity. Therefore, compared with existing technologies, this neuron circuit offers advantages such as simple structure and integrability, and can be applied in the field of neuromorphic computing. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The schematic diagram of the neuron circuit of N-type local active memristor;

[0012] Figure 2 is the small signal equivalent circuit of the N-type local active memristor;

[0013] Figure 3 is the frequency characteristic curve of the neuron circuit. DETAILED DESCRIPTION

[0014] The present invention will be further described below in conjunction with the accompanying drawings:

[0015] like Figure 1 Figure 2 shows the schematic diagram of a neuron circuit based on the proposed N-type local active memristor. The circuit consists of an N-type local active memristor, an inductor L, and an excitation voltage source VD. The negative electrode of the excitation voltage source VD is connected to ground, and the positive electrode is connected to one end of the inductor L. The other end of the inductor L is connected to one end of the N-type local active memristor, and the other end of the N-type local active memristor is grounded.

[0016] The process of establishing the mathematical model of the N-type locally active memristor includes the following steps:

[0017] Step 1: According to Chua's expansion theorem, the mathematical model of the general voltage-controlled memristor is:

[0018]

[0019] Where i and v represent the current flowing through the memristor and the voltage across the memristor; G M (x) is the memristor function; f(x,v) is a function of the memristor state variable x and voltage v.

[0020] Step 2: The present invention proposes a simple state variable equation:

[0021]

[0022] Among them, τ is used to change the rate of change of the memristor state variable, thereby determining the operating frequency range of the memristor.

[0023] Step 3: Let formula (3) equal to 0, so that the input voltage v = V, then

[0024] X=V 2 -αV (4)

[0025] Step 4: Substitute equation (4) into equation (2) to obtain the voltage-current relationship of the memristor at the operating point:

[0026]

[0027] Equation (5) describes the DC voltage-current characteristics of the N-type local active memristor. According to Equation (5), the differential conductance of the memristor at the DC operating point can be obtained as

[0028]

[0029] Step 5: In order to make the DC VI characteristic curve show N-type negative differential conductance characteristics and to simplify the state equation of the memristor as much as possible, the value of k should be at least equal to 1. In the present invention, k=1 in equation (6), and the derivation can be obtained.

[0030]

[0031] Where α=3(V1+V2) / 2, d0=3d1V1V2.

[0032] Step 6: According to equations (3) and (7), a mathematical model of an N-type local active memristor can be obtained:

[0033]

[0034] where τ = 10 -5 ,α=9,d1=10 -4 ,d0=2.4×10 -3 .

[0035] Using the small signal analysis method, the small signal impedance function of the N-type local active memristor can be obtained as:

[0036]

[0037] Where a1=1, a0=1 / τ, b1=d1X+d0, b0=(d1V·(2V-α)+d1X+d0) / τ. Therefore, the small signal equivalent circuit of the N-type local active memristor is shown in the attached figure. Figure 2 By connecting the N-type local active memristor in series with the inductor L, a neuron circuit based on the N-type local active memristor can be constructed. The circuit schematic is shown in the attached figure. Figure 1 As shown. Using Kirchhoff's law in the attached Figure 1 For the circuit shown, the circuit state equation is:

[0038]

[0039] The artificial neuron circuit is used to simulate the frequency characteristics of the tactile neurons in the biological sensory system. The inductance L is set to 300mH and the excitation voltage source V is set to D Increase the voltage from 2.1V to 3.85V in steps of 0.01V, and record the voltage at different V D The corresponding oscillation frequency value is plotted below, and the frequency change curve of the neuron circuit is shown in the attached figure. Figure 3 As shown in the figure, the oscillation frequency of the neuron circuit changes with the DC excitation V D The trend of increasing first and then decreasing is very similar to the "protective inhibition behavior" of biological tactile neurons.

Claims

1. A neuron circuit based on an N-type local active memristor, characterized by: The invention comprises an N-type local active memristor, an inductor L, and an excitation voltage source VD; the negative electrode of the excitation voltage source VD is connected to the ground, the positive electrode is connected to one end of the inductor L, the other end of the inductor L is connected to one end of the N-type local active memristor, and the other end of the N-type local active memristor is grounded; The mathematical model of the N-type local active memristor is: Where i, v, and x are the current flowing through the N-type local active memristor, the voltage across the N-type local active memristor, and the state variable of the N-type local active memristor, respectively. M is a memetic derivative function, d1, d0, τ, α are all constants.

2. A neuron circuit based on an N-type local active memristor according to claim 1, characterized in that: The τ=10 -5 ,α=9,d1=10 -4 ,d0=2.4×10 -3 .

Images