CN114648108A - Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron - Google Patents

Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron Download PDF

Info

Publication number
CN114648108A
CN114648108A CN202210418444.1A CN202210418444A CN114648108A CN 114648108 A CN114648108 A CN 114648108A CN 202210418444 A CN202210418444 A CN 202210418444A CN 114648108 A CN114648108 A CN 114648108A
Authority
CN
China
Prior art keywords
memristor
capacitor
volatile
volatile memristor
constant voltage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210418444.1A
Other languages
Chinese (zh)
Inventor
杨蕊
高森
缪向水
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Hubei Jiangcheng Laboratory
Original Assignee
Huazhong University of Science and Technology
Hubei Jiangcheng Laboratory
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology, Hubei Jiangcheng Laboratory filed Critical Huazhong University of Science and Technology
Priority to CN202210418444.1A priority Critical patent/CN114648108A/en
Publication of CN114648108A publication Critical patent/CN114648108A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • G06N3/065—Analogue means
    • G—PHYSICS
    • G11—INFORMATION STORAGE
    • G11C—STATIC STORES
    • G11C11/00—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
    • G11C11/54—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using elements simulating biological cells, e.g. neuron
    • G—PHYSICS
    • G11—INFORMATION STORAGE
    • G11C—STATIC STORES
    • G11C17/00—Read-only memories programmable only once; Semi-permanent stores, e.g. manually-replaceable information cards
    • G11C17/14—Read-only memories programmable only once; Semi-permanent stores, e.g. manually-replaceable information cards in which contents are determined by selectively establishing, breaking or modifying connecting links by permanently altering the state of coupling elements, e.g. PROM
    • G11C17/16—Read-only memories programmable only once; Semi-permanent stores, e.g. manually-replaceable information cards in which contents are determined by selectively establishing, breaking or modifying connecting links by permanently altering the state of coupling elements, e.g. PROM using electrically-fusible links

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Neurology (AREA)
  • Theoretical Computer Science (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Hardware Design (AREA)
  • Logic Circuits (AREA)

Abstract

The invention discloses a memristor-based self-adaptive bionic neuron circuit and a method for realizing self-adaptive simulation of a bionic neuron by using the same, wherein an excitation pulse and a capacitor C1 form a first charging loop; the volatile memristor M1, the constant voltage source V1 and the capacitor C1 form a first reverse charging loop; the volatile memristor M1, the constant voltage source V1 and the capacitor C2 form a second reverse charging loop; the volatile memristor M2, the constant voltage source V2, and the capacitor C2 constitute a second charging loop. The method has the advantages that the volatile memristor is subjected to threshold value transition behavior by utilizing the capacitor charging behavior, the basic function of generating action potential by the neuron is realized by adding the continuous output of the constant voltage source, the threshold voltage of the volatile memristor gradually changes in the working process of the excitation pulse, the multi-mode action potential distribution similar to that of the biological neuron under constant excitation and the change of the discharge frequency are realized, and the neuron self-adaption capability is realized.

Description

Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron
Technical Field
The invention belongs to the technical field of brain-like bionic technology, and particularly relates to a memristor-based self-adaptive bionic neuron circuit and a bionic neuron self-adaptive simulation method.
Background
Computing systems based on the von neumann architecture are prone to performance bottlenecks in the face of large-scale data processing due to memory wall limitations. On the contrary, the human brain has the capability of rapidly processing a large amount of information in parallel, and can realize the integration of information processing and storage. There are billions of different types of neurons distributed in the human brain, which are interconnected by millions of synapses between them, supporting the brain's processing and storing information. Therefore, the hardware is used for constructing the basis of a brain-like chip, and two basic units, namely an artificial neuron and an artificial synapse, are designed.
Memristors were proposed and implemented as milestones for brain-like computing related research. At present, researchers have built artificial Leaky-Integrated-and-fire (LIF) neurons by using simple circuits and memristors. However, it is worth noting that since the LIF neuron model only considers the threshold firing characteristics of neurons, it can only simulate the change of neuron firing frequency by changing the resistance magnitude of the input terminal coupling. In order to realize the function of a more bionic biological neuron, a Hodgkin-Huxley (HH) neuron model is needed to simulate the multi-mode action potential firing similar to the biological neuron and the change of the discharge frequency under constant excitation, namely, the neuron self-adaptive capacity is realized, which is the direction of the effort needed at present.
Disclosure of Invention
In view of the above defects or improvement requirements of the prior art, the present invention provides a memristor-based adaptive bionic neuron circuit and a bionic neuron adaptive simulation method, which aim to realize simulation of multi-mode action potential firing similar to a biological neuron and change of discharge frequency under constant excitation, that is, to realize neuron adaptive capability.
To achieve the above object, according to one aspect of the present invention, there is provided a memristor-based adaptive bionic neuron circuit, including a volatile memristor M1, a volatile memristor M2, a constant voltage source V1, a constant voltage source V2, a capacitor C1, a capacitor C2, and a resistor R1, wherein threshold voltages of the volatile memristor M1 and the volatile memristor M2 are adjustable, and,
two ends of the capacitor C1 are used for receiving the excitation pulse;
the negative end of the volatile memristor M1 is connected with the negative electrode of the constant voltage source V1, the positive end of the volatile memristor M1 is connected to the first end of the capacitor C1, and the positive electrode of the constant voltage source V1 is connected to the second end of the capacitor C1;
the positive end of the volatile memristor M2 is connected with the positive electrode of the constant voltage source V2, the negative end of the volatile memristor M2 is connected to the first end of the capacitor C2, and the negative electrode of the constant voltage source V2 is connected to the second end of the capacitor C2;
the first end of the capacitor C2 is connected to the first end of the capacitor C1 through the resistor R1, the second end of the capacitor C2 is connected to the second end of the capacitor C1, and both ends of the capacitor C2 are used as circuit output ends to generate an action potential.
In one embodiment, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 change with the change of the environment or the self material structure.
In one embodiment, the materials of the volatile memristor M1 and the volatile memristor M2 comprise VO2。
In one embodiment, the system further comprises a resistor R2, a resistor R3 and a resistor R4, wherein,
the positive terminal of the volatile memristor M1 is connected with the first terminal of the capacitor C1 through the resistor R2;
the negative end of the volatile memristor M2 is connected with the first end of the capacitor C2 through a resistor R3;
the capacitor C1 is connected to pulse excitation through the resistor R4.
In one embodiment, the constant voltage provided by the constant voltage source V1 is less than the threshold voltage of the volatile memristor M1, and the constant voltage provided by the constant voltage source V2 is less than the threshold voltage of the volatile memristor M2.
In one embodiment, the volatile memristor M1 and the volatile memristor M2 are bipolar memristors or unipolar memristors.
In one of the embodiments, further comprising,
a threshold voltage regulation device to regulate threshold voltages of the volatile memristor M1 and the volatile memristor M2 during receipt of the stimulation pulse.
According to another aspect of the present invention, there is provided a method for bionic neuron adaptive simulation, comprising:
building the adaptive bionic neuron circuit based on the memristor;
an excitation pulse is applied to two ends of the capacitor C1, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 are gradually adjusted, and the pulse frequency output by the output end of the circuit changes along with the change of the threshold.
In one embodiment, the pulse output by the circuit output is switched between a unimodal oscillation pulse mode and a multi-modal cluster pulse mode.
In one embodiment, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 change along with the change of the environment or the material structure of the volatile memristor M1 and the volatile memristor M2, and the environment or the material structure of the volatile memristor M1 and the volatile memristor M2 gradually changes during the application of the excitation pulse to the two ends of the capacitor C1.
In the application, a specific adaptive bionic neuron circuit is built based on a memristor, wherein an excitation pulse and a capacitor C1 form a first charging loop; the volatile memristor M1, the constant voltage source V1 and the capacitor C1 form a first reverse charging loop; the volatile memristor M1, the constant voltage source V1 and the capacitor C2 form a second reverse charging loop; the volatile memristor M2, the constant voltage source V2 and the capacitor C2 form a second charging loop; the voltage across the capacitor C2 generates an action potential as an output pulse. Based on the built adaptive bionic neuron circuit, after an excitation pulse is applied, the capacitor C1 is charged through the first charging loop, the voltage of the capacitor C1 is gradually increased, when the voltage across the volatile memristor M1 is larger than or equal to a threshold voltage, the volatile memristor M1 is changed into a low-resistance state, the constant voltage source V1 reversely charges the capacitor C1 through the first reverse charging loop and charges the capacitor C2 through the second reverse charging loop, the voltages of the capacitor C1 and the capacitor C2 are quickly reduced, when the voltage across the volatile memristor M2 is increased to be larger than or equal to the threshold voltage, the volatile memristor M2 is changed into the low-resistance state, the constant voltage source V2 charges the capacitor C2 through the second charging loop, and the voltage across the capacitor C2 serves as an output end to generate an action potential. Meanwhile, as the threshold voltages of the volatile memristor M1 and the volatile memristor M2 are adjustable, in the working process of the excitation pulse, the thresholds of the volatile memristors M1 and M2 are constantly and regularly changed, or constantly increased or constantly decreased, and the charging and discharging time of the capacitor is constantly changed, so that the generated action potential frequency and the issuing mode are changed, and the self-adaption function of the simulated neuron is realized.
Drawings
FIG. 1 is a circuit schematic of an adaptive biomimetic neuron based on memristors of an embodiment;
FIG. 2 shows an example of a volatile memristor at a normal temperature t0To a preset high temperature thighSchematic diagram of I-V characteristics;
FIG. 3 is a schematic diagram of an embodiment of constant voltage pulse excitation;
FIG. 4 shows an embodiment at a predetermined high temperature thighThe lower neuron circuit generates an output schematic diagram of action potential;
FIG. 5 is a schematic diagram of an embodiment of an output of a neuron circuit generating an action potential at a normal temperature;
FIG. 6 shows an embodiment of a neuron circuit under constant electrical pulse excitation, VthThe adaptive function schematic diagram under the condition of continuous reduction;
FIG. 7 shows an embodiment of a neuron circuit under constant electrical pulse excitation, VthThe adaptive function schematic diagram under the condition of continuous increase.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
FIG. 1 is a schematic diagram of an adaptive bionic neuron circuit based on memristors in an embodiment. The circuit specifically includes a volatile memristor M1, a volatile memristor M2, a constant voltage source V1, a constant voltage source V2, a capacitance C1, a capacitance C2, and a resistance R1.
Wherein, two ends of the capacitor C1 are used for receiving the excitation pulse. Specifically, the pulse generator for generating the excitation pulse may be a part of the circuit, or may be separated from the circuit, and the excitation pulse may be input to the capacitor C1 from outside the circuit. The excitation pulse may be a voltage pulse or a current pulse. The capacitor C1 and the excitation pulse form a first charging loop for charging the capacitor C1. Specifically, a resistor R4 may be disposed in the first charging loop, and the capacitor C1 may be connected to the excitation pulse through the resistor R4.
The negative end of the volatile memristor M1 is connected with the negative electrode of the constant voltage source V1, the positive end of the volatile memristor M1 is connected to the first end of the capacitor C1, and the positive electrode of the constant voltage source V1 is connected to the second end of the capacitor C1. The volatile memristor M1, the constant voltage source V1, and the capacitance C1 form a first reverse charging loop. Specifically, a resistor R2 may be disposed within the first reverse charging loop, the resistor R2 being located between the first end of the capacitor C1 and the positive end of the volatile memristor M1.
The positive end of the volatile memristor M2 is connected with the positive electrode of the constant voltage source V2, the negative end of the volatile memristor M2 is connected to the first end of the capacitor C2, and the negative electrode of the constant voltage source V2 is connected to the second end of the capacitor C2. The volatile memristor M2, the constant voltage source V2, and the capacitor C2 form a second charging loop. Specifically, a resistor R3 may be disposed in the second reverse charging loop, and the resistor R3 is located between the negative terminal of the volatile memristor M2 and the first terminal of the capacitor C2. It should be noted that the resistance state of the memristor changes when the positive terminal of the memristor receives a forward stimulus. Specifically, the volatile memristors M1 and M2 may be unipolar memristors or bipolar memristors. When the volatile memristor is a unipolar memristor, the positive end and the negative end of the unipolar memristor are fixed and irreversible; when the volatile memristor is a bipolar memristor, any one of the two ends can be selected as a positive end, and the other end can be selected as a negative end.
The first end of the capacitor C2 is connected to the first end of the capacitor C1 through the resistor R1, and the second end of the capacitor C2 is connected to the second end of the capacitor C1. The volatile memristor M1, the constant voltage source V1, and the capacitance C2 constitute a second reverse charging loop. The two ends of the capacitor C2 are used as circuit output ends to generate an action potential. Specifically, the second terminal of the capacitor C1, the positive terminal of the constant voltage source V1, the negative terminal of the constant voltage source V2, and the second terminal of the capacitor C2 are all grounded.
The resistance states of the above volatile memristors M1 and M2 vary with voltage. The initial states of the volatile memristors M1 and M2 are high-resistance states, in the process that the voltages of the two ends of the volatile memristors M1 and M2 are increased from 0 to above, when the voltages of the two ends of the volatile memristors M1 and M2 are smaller than the threshold voltages of the volatile memristors, the volatile memristors are kept in the high-resistance states, and when the voltages of the two ends of the volatile memristors M1 and M2 are larger than or equal to the threshold voltages of the volatile memristors, the volatile memristors M1 and M2 are converted into the low-resistance states; after the volatile memristors M1 and M2 are converted into the low-resistance state, the low-resistance state can be kept all the time under the condition that the voltage at the two ends of the low-resistance state is larger than or equal to the holding voltage, the voltage at the two ends of the volatile memristors M1 and M2 is reduced to be smaller than the holding voltage, and the volatile memristors M1 and M2 recover to the initial high-resistance state. The holding voltage of the volatile memristors M1, M2 is less than their threshold voltages.
Based on the self-adaptive bionic neuron circuit, the integration and release processes of the neuron can be realized, and the specific processes are as follows:
the initial resistance states of the volatile memristors M1 and M2 are high resistance states, after the first end of the resistor R4 receives an excitation pulse, the capacitor C1 is charged through the first charging loop, the voltage at two ends of the capacitor C1 is gradually increased, the positive end voltage of the volatile memristor M1 is also increased, but the high resistance state of the volatile memristor M1 is kept unchanged, and the process is a neuron integration process.
When the voltage across the volatile memristor M1 is greater than or equal to the threshold voltage, the volatile memristor M1 is rapidly changed from the high-resistance state to the low-resistance state, the constant voltage source V1 charges the capacitor C1 through the first reverse charging loop and charges the capacitor C2 through the second reverse charging loop, the voltages of the capacitor C1 and the capacitor C2 are rapidly reduced, the voltage of the negative terminal of the volatile memristor M2 is reduced, and the voltage difference across the volatile memristor M2 is increased; when the voltage across the volatile memristor M2 is greater than or equal to the threshold voltage, the volatile memristor M2 becomes a low-resistance state, the constant voltage source V2 charges the capacitor C2 through the second charging loop, the voltage across the capacitor C2 increases rapidly, the voltage across the capacitor C2 serves as an output pulse to generate an action potential, and the process is a neuron release process. The overall implementation achieves a neuron basis integrate-and-release function.
In the present application, the volatile memristors M1, M2 have an initial resistance state that is a high resistance state. At both ends of which the voltage increases from 0 to a threshold voltage VthThe device remains in the high resistance state. At a voltage across it greater than a threshold voltage VthThe device then quickly transitions from the high resistance state to the low resistance state. At both ends of which the voltage is gradually reduced to a holding voltage VholdBefore, the device keeps the low resistance state unchanged, and when the voltage across the device is reduced to be lower than the holding voltage VholdThe device then reverts from the low resistance state to the high resistance state. Wherein the threshold voltage VthGreater than the holding voltage Vhold。
In the present application, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 are adjustable. In the working process of the excitation pulse, the threshold values of the volatile memristors M1 and M2 are changed regularly or are increased or reduced continuously, so that the charging time required for the volatile memristors M1 and M2 to generate resistance state transition in the neuron circuit is changed under the constant excitation pulse, and finally, the action potential frequency and the issuing mode generated by the circuit are changed to simulate the adaptive function of the neuron.
For stable operation of the circuit, in one embodiment, the constant voltage supplied by the constant voltage source V1 is required to be less than the threshold voltage of the volatile memristor M1, and the constant voltage supplied by the constant voltage source V2 is required to be less than the threshold voltage of the volatile memristor M2.
In one embodiment, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 change with the change of the environment or the material structure of the volatile memristor M, so that the output of the circuit changes along with the change of the environment or the material structure of the volatile memristor M2 during the operation of the circuit, and the adaptive function of a neuron is simulated. The environmental change can be temperature, voltage and other changes, namely the threshold voltage of the volatile memristor can change along with the environmental temperature and can also be regulated and controlled by an electric field. For example, the volatile memristor is designed to be similar to a three-terminal structure of a transistor, electrodes at two ends of a source and a drain are connected to a circuit to work, when the gate voltage is set to be different values, an electric field effect is generated on the memristor material, the threshold voltage of the memristor is influenced, and the larger the gate voltage is, the smaller the threshold voltage of the memristor is. In some embodiments, the threshold voltage of the volatile memristor can be changed by adjusting the internal material structure of the volatile memristor. For example, by applying a voltage across the device, a metal conductive bridge mechanism of a volatile memristor is utilized to generate a redox reaction and migration of active metal ions at an active metal electrode to form a conductive wire, so that the memristor device completes the transition from a high-resistance state to a low-resistance state, the conductive wire needs to be reversibly generated and broken by applying the voltage across the device, and each generation and breaking of the conductive wire strengthens the formation of a channel, which makes the next generation easier, namely, the threshold voltage of the memristor is gradually reduced.
Specifically, the memristor can select NbO2A chalcogenide phase change material such as GeTe, in this embodiment, VO is selected2。
With VO2For illustration. According to the theoretical analysis of energy bands, VO of rutile structure2(R) VO externally exhibiting a metallic state and a monoclinic lattice2(M) exhibits a semiconductor state to the outside. Through experimental research and analysis, the VO can be excited by stimulation of heat, electricity, light and the like2A phase transition occurs. At normal temperature, VO2Is a semiconductor state, corresponding to the high resistance state of the volatile memristors M1 and M2, VO is generated when heat is used as a stimulus2When the temperature is increased from the room temperature to around 68 ℃, the monoclinic system is transformed to the tetragonal system, namely, the phase transformation is carried out to be the metal state, which corresponds to the low resistance state of the volatile memristors M1 and M2. When VO is present2In the process of cooling from the high-temperature state to the room temperature when the phase change occurs, a metal-semiconductor conversion process occurs and the semiconductor state is returned again. Also, we can use electrical stimulation instead of heat to make VO2For the field of volatile memristors, where the voltage across it increases from 0 to a threshold voltage VthIn the process, the device maintains the semiconductor state and thus exhibits a high resistance state. At a voltage across it greater than a threshold voltage VthThe device then undergoes a phase transition from the semiconductor state to the metal state. At both ends of which the voltage is gradually reduced to a holding voltage VholdThe device then spontaneously returns to the semiconductor state, from the low resistance state to the high resistance state. Wherein the threshold voltage VthGreater than a holding voltage Vhold. In addition, the threshold voltage is also reduced as the ambient temperature is higher, and fig. 2 shows that the volatile memristors M1 and M2 are at the normal temperature t0To a preset high temperature thighThe I-V characteristic diagram at (45 ℃) shows that the threshold voltage of the volatile memristor is reduced along with the increase of the temperature.
The capacitors C1 and C2 are fixed capacitors or variable capacitors, and have capacitance values ranging from 10fF to 10 μ F, in this embodiment, the value of the capacitor C1 is 16nF, and the value of the capacitor C2 is 4 nF. The resistance is a constant value resistance or a variable resistance, the resistance value is smaller than the off-state resistance of the volatile memristor, in the embodiment, the resistance R2 is 120 Ω, the resistance R3 is 0, the resistance R4 is 1k Ω, and the resistance R1 is 1k Ω. The voltage magnitude of the constant voltage sources V1, V2 is less than the threshold voltage of the volatile memristor, and in this embodiment, the voltage magnitude of the constant voltage sources V1, V2 is 1.95V. The excitation pulse is a current pulse or a voltage pulse, and in the present embodiment, the excitation pulse is a voltage pulse. Fig. 3 is a schematic diagram of constant voltage pulse excitation according to an embodiment of the present invention. FIG. 4 is a schematic diagram of the input/output of the neuron circuit for generating an action potential (preset high temperature t)high) The action potential is in an oscillation emission mode, and the frequency of the action potential is about 1.5 x 105hz; FIG. 5 is a schematic diagram of the input/output of the neuron circuit generating the action potential according to the embodiment of the present invention (normal temperature t)0) The action potential is in a cluster emission mode, and a single cluster contains two peak valuesAn action potential with an action potential frequency of 6.7 x 104hz。
In an embodiment, the adaptive bionic neuron circuit may be further configured with a threshold voltage regulation device for regulating the threshold voltages of the volatile memristor M1 and the volatile memristor M2 during receipt of an excitation pulse. For example, a temperature control device may be arranged to adjust the temperature of the environment in which the volatile memristor M1 and the volatile memristor M2 are located, so that the threshold voltages of the volatile memristors M1 and M2 change with the change of the environment temperature.
Correspondingly, the application also relates to a bionic neuron self-adaptive simulation method, which utilizes the self-adaptive bionic neuron circuit based on the memristor introduced above to apply excitation pulses to two ends of the capacitor C1, and gradually adjusts the threshold voltages of the volatile memristor M1 and the volatile memristor M2, so that the pulse frequency output by the output end of the circuit changes along with the change of the threshold.
More specifically, the pulse output by the circuit output end is switched between a unimodal oscillation pulse mode and a multi-modal cluster pulse mode, and the unimodal oscillation pulse mode and the multi-modal cluster pulse mode are matched with a real biological neuron model, so that the adaptive function of a simulated neuron can be better simulated.
More specifically, the threshold voltages of the volatile memristors M1 and M2 change with the change of the environment or the material structure of the volatile memristors M1 and M2, and the environment or the material structure of the volatile memristors M1 and M2 gradually changes during the application of the excitation pulse to the two ends of the capacitor C1.
The neuron circuit is capable of simulating neuron adaptation functions. Taking the example that the applied excitation pulse is a voltage pulse and the threshold voltage is changed along with the temperature, the method has two operation modes:
1) the initial working environment temperature of the neuron circuit is set to be normal temperature, as the operation of the neuron circuit is carried out, each element of the circuit generates heat due to power consumption during operation, the working temperature of each element is increased continuously, the threshold voltage of the volatile memristors M1 and M2 is reduced gradually, the charging time required by the volatile memristors M1 and M2 to generate resistance state transition in the neuron circuit under a constant excitation pulse is reduced gradually, finally the action potential frequency generated by the circuit is increased rapidly, and the firing mode is transmitted from a cluster to the oscillation transition, as shown in FIG. 6, so that the self-adaptive function of a neuron is simulated.
2) The initial working environment temperature of the neuron circuit is set to be a preset high temperature, the temperature control device is turned off after the neuron circuit starts to work, as the work of the neuron circuit progresses, the heat dissipation capacity of each element of the circuit is larger than the heat generation capacity of power consumption, the working temperature of each element of the circuit approaches to the room temperature, the threshold voltage of the volatile memristors M1 and M2 gradually increases, the charging time required by the volatile memristors M1 and M2 to generate resistance state transition in the neuron circuit under constant excitation pulse gradually increases, finally the frequency of the action potential generated by the circuit is rapidly reduced, and the transition from oscillation to cluster emission is carried out in the emitting mode, as shown in figure 7, and therefore the self-adaption function of the neuron is simulated.
According to the neuron circuit, threshold transition behaviors of volatile memristors M1 and M2 are achieved through charging behaviors of capacitors C1 and C2, continuous output of constant voltage sources V1 and V2 is added, the basic function of generating action potentials of neurons is achieved, in the working process of excitation pulses, the threshold voltages of the volatile memristors M1 and M2 gradually change regularly, or continuously increase or continuously decrease, multi-mode action potential emitting and discharging frequency changing similar to those of biological neurons under constant excitation are achieved, and therefore the neuron self-adaption capability is achieved.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (10)

1. An adaptive bionic neuron circuit based on memristors is characterized by comprising a volatile memristor M1, a volatile memristor M2, a constant voltage source V1, a constant voltage source V2, a capacitor C1, a capacitor C2 and a resistor R1, wherein threshold voltages of the volatile memristor M1 and the volatile memristor M2 are adjustable, and,
two ends of the capacitor C1 are used for receiving the excitation pulse;
the negative end of the volatile memristor M1 is connected with the negative electrode of the constant voltage source V1, the positive end of the volatile memristor M1 is connected to the first end of the capacitor C1, and the positive electrode of the constant voltage source V1 is connected to the second end of the capacitor C1;
the positive end of the volatile memristor M2 is connected with the positive electrode of the constant voltage source V2, the negative end of the volatile memristor M2 is connected to the first end of the capacitor C2, and the negative electrode of the constant voltage source V2 is connected to the second end of the capacitor C2;
the first end of the capacitor C2 is connected to the first end of the capacitor C1 through the resistor R1, the second end of the capacitor C2 is connected to the second end of the capacitor C1, and both ends of the capacitor C2 are used as circuit output ends to generate an action potential.
2. The memristor-based adaptive biomimetic neuron circuit as recited in claim 1, wherein threshold voltages of the volatile memristor M1 and the volatile memristor M2 vary with the environment or the material structure thereof.
3. The memristor-based adaptive biomimetic neuron circuit of claim 2, wherein the materials of volatile memristor M1 and volatile memristor M2 comprise VO2。
4. The memristor-based adaptive biomimetic neuron circuit of claim 1, further comprising a resistance R2, a resistance R3, and a resistance R4, wherein,
the positive terminal of the volatile memristor M1 is connected with the first terminal of the capacitor C1 through the resistor R2;
the negative end of the volatile memristor M2 is connected with the first end of the capacitor C2 through a resistor R3;
the capacitor C1 is connected to pulse excitation through the resistor R4.
5. The memristor-based adaptive biomimetic neuron circuit as claimed in claim 1, wherein the constant voltage provided by the constant voltage source V1 is less than a threshold voltage of the volatile memristor M1, and the constant voltage provided by the constant voltage source V2 is less than a threshold voltage of the volatile memristor M2.
6. The memristor-based adaptive biomimetic neuron circuit as in claim 1, wherein the volatile memristor M1 and the volatile memristor M2 are bipolar memristors or unipolar memristors.
7. The memristor-based adaptive bionic neuron circuit according to any one of claims 1 to 6, further comprising,
a threshold voltage regulation device to regulate threshold voltages of the volatile memristor M1 and the volatile memristor M2 during receipt of the stimulation pulse.
8. A bionic neuron adaptive simulation method is characterized by comprising the following steps:
building an adaptive biomimetic memristor-based neuron circuit as defined in any one of claims 1 to 6;
an excitation pulse is applied to two ends of the capacitor C1, the threshold voltages of the volatile memristor M1 and the volatile memristor M2 are gradually adjusted, and the pulse frequency output by the output end of the circuit changes along with the change of the threshold.
9. The method of adaptive simulation of biomimetic neurons according to claim 8, wherein the pulse output by the circuit output is switched between a unimodal oscillatory pulse and a multi-modal cluster pulse.
10. The bionic neuron adaptive simulation method as claimed in claim 8, wherein threshold voltages of the volatile memristor M1 and the volatile memristor M2 change along with changes of the environment or the material structure of the volatile memristor M1, and the environment or the material structure of the volatile memristor M1 and the volatile memristor M2 gradually changes during application of excitation pulses to two ends of the capacitor C1.
CN202210418444.1A 2022-04-20 2022-04-20 Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron Pending CN114648108A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210418444.1A CN114648108A (en) 2022-04-20 2022-04-20 Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210418444.1A CN114648108A (en) 2022-04-20 2022-04-20 Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron

Publications (1)

Publication Number Publication Date
CN114648108A true CN114648108A (en) 2022-06-21

Family

ID=81996510

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210418444.1A Pending CN114648108A (en) 2022-04-20 2022-04-20 Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron

Country Status (1)

Country Link
CN (1) CN114648108A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115169507A (en) * 2022-09-08 2022-10-11 华中科技大学 Brain-like multi-mode emotion recognition network, recognition method and emotion robot
CN115906961A (en) * 2023-02-22 2023-04-04 北京大学 Self-adaptive artificial pulse neuron circuit based on volatile threshold resistance changing memristor

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115169507A (en) * 2022-09-08 2022-10-11 华中科技大学 Brain-like multi-mode emotion recognition network, recognition method and emotion robot
CN115906961A (en) * 2023-02-22 2023-04-04 北京大学 Self-adaptive artificial pulse neuron circuit based on volatile threshold resistance changing memristor

Similar Documents

Publication Publication Date Title
CN114648108A (en) Self-adaptive bionic neuron circuit and self-adaptive simulation method of bionic neuron
CN103078054B (en) Unit, device and method for simulating biological neuron and neuronal synapsis
CN107742153B (en) Memristor-based neuron circuit with steady-state plasticity
CN103580668B (en) A kind of associative memory circuit based on memristor
CN103430186A (en) Learning method for neural network circuit
Nishitani et al. Dynamic observation of brain-like learning in a ferroelectric synapse device
CN110428050B (en) Synapse bionic circuit for realizing diversified STDP learning rules based on memristor
US20100299297A1 (en) System for electronic learning synapse with spike-timing dependent plasticity using phase change memory
Chen et al. Associate learning and correcting in a memristive neural network
CN103460220A (en) Neural network circuit learning method
CN210488595U (en) Synapse bionic circuit for realizing diversified STDP learning rule based on memristor
Mahalanabis et al. Demonstration of spike timing dependent plasticity in CBRAM devices with silicon neurons
CN112906880B (en) Adaptive neuron circuit based on memristor
Kong et al. Study of recall time of associative memory in a memristive Hopfield neural network
Zheng et al. Memristor-based synapses and neurons for neuromorphic computing
JP6989552B2 (en) Neural network device
CN114970840A (en) Irregular discharge bionic neuron circuit and discharge bionic method thereof
Wu et al. Threshold switching memristor-based voltage regulative circuit
Wu et al. Bursting dynamics in a spiking neuron with a memristive voltage-gated channel
US11922298B2 (en) Neuron device using spontaneous polarization switching principle
KR20210088444A (en) On-chip training neuromorphic architecture
Singha et al. Analog memristive time dependent learning using discrete nanoscale RRAM devices
CN108777153B (en) Multi-terminal input synapse device and plasticity modulation method thereof
Howard et al. A spice model of the peo-pani memristor
Shin et al. Neuronal spike event generation by memristors

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination