CN106845634A - A kind of neuron circuit based on memory resistor - Google Patents
A kind of neuron circuit based on memory resistor Download PDFInfo
- Publication number
- CN106845634A CN106845634A CN201611235356.9A CN201611235356A CN106845634A CN 106845634 A CN106845634 A CN 106845634A CN 201611235356 A CN201611235356 A CN 201611235356A CN 106845634 A CN106845634 A CN 106845634A
- Authority
- CN
- China
- Prior art keywords
- memristor
- cynapse
- neuron
- resistance
- neuron circuit
- 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.)
- Granted
Links
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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Neurology (AREA)
- Semiconductor Memories (AREA)
Abstract
The invention discloses a kind of neuron circuit based on memory resistor, in the present invention, the memristor of cynapse array selects part volatibility bipolarity electric resistance changing device, and the memristor for expressing membrane potential of neurons selects volatibility electric resistance changing device, neuron circuit is built, and with cynapse elementary cell.The neuron circuit can realize the integration discharging function in biological neuron, give expression to local hierarchical current potential, cynapse has part volatibility, can express the related plasticity of activity schedule, has great similarity in terms of information Store, transmission and treatment with biologically neuron and cynapse.The present invention can provide elementary cell for hardware simulation cerebral nerve network structure, the Neural spike train time delay for overcoming prior art to exist, it is difficult to the technical problems such as High Density Integration, the information processing system of class brain can be used to construct, bulk information can be quickly processed parallel to have very big application value in the neurology calculating network for realizing brain.
Description
Technical field
The invention belongs to semiconductor message area, and in particular to a kind of neuron circuit based on memory resistor.The circuit
It is a kind of neuron for artificial neural network and cynapse elementary cell, it is deposited with biologically neuron and cynapse in information
Storage, transmission and treatment aspect have great similarity, may be used to build the neural computing network of class brain.
Background technology
The brain of people is better than contemporary most computers at aspects such as cognitive function, language understanding, abstract reasonings, while tool
There is the features such as small volume, low in energy consumption, efficiency high, fault tolerant concurrent operation.Traditional computer is based on von Karman structure
, information processing is performed separately with storage, and concurrent operation is indifferent.It is different from computer, in big intracerebral, the treatment of information with
Storage is carried out in the same time and place.Human brain is by about 1011Individual neuron passes through about 1015Individual cynapse is connected with each other, and forms one
Huge neutral net, can quickly process bulk information parallel.
Neuron serves key effect in brain information processing procedure, and the major function of neuron is to process and transmit
Information, and complete the cell membrane that this function relies primarily on neuronal cell.In the brain, neuron receives from the emerging of dendron
Putting forth energy property or inhibitory synapse current potential, producing has certain ageing local hierarchical current potential, and is integrated.Lipid bilayer
The current potential of cell membrane then can accordingly change, and when certain value is reached, neuron can then produce action potential, send signal, and
Signal is delivered to next neuron by aixs cylinder via cynapse.It is one of most basic function of neuron to integrate electric discharge.God
Key effect is also played through cynapse (two connecting portions of neuron), the form and function of its plasticity, i.e. cynapse are by the external world
The influence of stimulation and the characteristic that changes are brain learnings and memory, the physiological Foundations of information processing and storage.Therefore,
The key that artificial neural network is built from hardware is to develop the artificial neuron with class nervous function, and the work(with cynapse
Energy.
The neuron realized using traditional cmos circuit is, it is necessary to use the transistor AND gate electric capacity of complexity, and be difficult to and high
Density cynapse array is integrated, also, tradition silicon-based transistor neuron is mainly used in Digital Logical Circuits at present, its pulse nerve
Function is simultaneously not implemented.And the special circuit for simulating a synaptic function is accomplished by tens triodes, and the neutral net of human brain
In there are about 1015Individual cynapse.Therefore, setting up the huge neutral net suitable with human brain on hardware based on traditional cmos circuit is
It is unpractical.
Research on memristor in the last few years shows, the memristor with simple sandwich structure, in the work of electric pulse
With the lower continuous enhancing that can realize conduction and reduction, for imictron and cynapse basic function, thus receive extensively
Ground concern.First, the electric conductivity that memristor is gradually changed has great similarity with the plasticity of biology cynapse.Memory resistor
Electric conductivity enhancing can simulate the enhancing of biology synaptic connection strengths;Memory resistor electric conductivity reduces can simulate Synaptic junction
The suppression of intensity.And neuron is when processing and transmitting electric signal, cell membrane potential is also continuous accumulation, i.e. integration process.
Thus memristor can simulate the change of film potential as key element, with reference to other elements, realize the integration electric discharge of neuron
Function.
Resistance states currently used for the memristor of analog neuron cynapse are completely non-volatile, i.e., its electric conductivity is outside
Added electric field is maintained at certain numerical value after removing, do not change over time.However, the bonding strength of cynapse is in telecommunications in biology
Number effect after, can first strengthen, gradually decay to a certain extent then as the time.This of synaptic plasticity dynamically becomes with the time
The process of change, can realize the time encoding to bursting activity, to realizing that it is important that the functions such as brain learning, memory, forgetting have
Meaning.
And be based on memristor structure neuron circuit report it is very few, be badly in need of the problem to be overcome.
Memristor simple structure, is designed by cross spider, can High Density Integration.Thus, neuron is built by memristor
Circuit, and with synaptic function, it is significant.Meanwhile, the device for expressing synaptic plasticity uses part volatibility memristor, more
The nearly biology cynapse characteristic of adjunction, this neuron will have major application prospect in Artificial Intelligence Circuits.
The content of the invention
The present invention proposes a kind of neuron circuit built based on memory resistor, is used to realize the plasticity of synapse
With the simulation that membrane potential of neurons integrates flash-over characteristic, the Neural spike train time delay for overcoming prior art to exist, it is difficult to real
The technical problems such as existing High Density Integration.
Neuron circuit includes cynapse array, dendron, pericaryon and aixs cylinder.Cynapse is used for receiving and adjusting upper level
The action potential that neuron circuit occurs, pericaryon is delivered to through dendron, in the case where being delivered to through aixs cylinder after integrating electric discharge again
One-level neuron circuit.Dendron, is realized by connecting wire;Aixs cylinder, is realized by connecting wire;Cynapse part, by easy with part
The memristor of the property lost is realized;Pericaryon, by complete volatibility or the memristor of the expression membrane potential of neurons of part volatibility
Device and other corresponding electronic device devices are realized.
Conceive based on more than, concrete technical scheme of the present invention is as follows:
A kind of neuron circuit based on memory resistor, including cynapse array, switching tube, adder, expression neuron membrane
The memristor of current potential, comparator, Spike signal generators;Wherein:
The cynapse array is used to receive the action potential that upper level neuron circuit is transmitted, and it includes that several are arranged side by side
Part volatibility memristor, (in the present invention, aixs cylinder is by leading for each aixs cylinder of each memristor one end and upper level neuron circuit
Line is realized) it is connected, other end tandem is all the way, to be connected with adder input through switch;
The adder is used to integrate the action potential of each input of cynapse array, for adjusting expression membrane potential of neurons
Memristor resistance, so as to realize the simulation to membrane potential of neurons;
The memristor of the expression membrane potential of neurons is complete volatibility memristor, for imictron cell membrane
Local hierarchical current potential;One terminates the output end of the adder, and the other end is divided into two-way, and divider resistance ground connection is connect all the way;Separately
Comparator is connect all the way, will deliver to comparator input terminal after action potential partial pressure after the integration of acquisition;
Another input termination reference voltage V of comparatorR, for comparing divider resistance voltage-to-ground and reference voltage VR
Size;When divider resistance voltage-to-ground is more than VRWhen, conduction level is exported, otherwise export cut-off level;
The Spike signal generators input is connected with comparator output terminal, and Spike signal generators export three roads letter
Number, it is connected with the control pole with the switching tube all the way, for the action of controlling switch pipe;Second road signal connects cynapse array
Tandem end, the transmission efficiency for adjusting cynapse array;3rd road signal connects next stage neuron circuit, used as next stage nerve
First circuit input signal.Usually during comparator output cut-off level, Spike signal generators are stopped, and turn on switching tube;
When comparator exports conduction level, Spike signal generators export a cut-off signals, disconnect switching tube;
During work, the upper level neuron signal being input into via cynapse array is added by adder, enters horizontal electrical signal reality
When integrate, electric signal amplitude rises to threshold value (this value depends on the memristor of expression membrane potential of neurons used after integration
Resistance switching performance, those skilled in the art know that electric resistance changing device has corresponding threshold value) when, make expression membrane potential of neurons
Memristor resistance reduce so that divider resistance voltage-to-ground increase;When divider resistance voltage-to-ground exceedes reference voltage VR
When, comparator output conduction level, Spike signal generators are realized according to the biological electric signal electric discharge of the simulation for prestoring
Sending for action potential, completes the integration discharging function of neuron;The reference voltage VRIt is according to expression membrane potential of neurons
The resistance of memristor and the resistance size of divider resistance and neuron action potential amplitude mutually weigh depending on set
One constant pressure.
Further, each cynapse in the cynapse array, is realized using part volatibility memristor.
Further, the operational amplifier that the adder is used.
Further, the memristor of the expression membrane potential of neurons is from part volatibility device or completely non-volatile
Device, Spike signal generators output end is connected with the memristor of expression membrane potential of neurons and the output end of adder, is used for
When neuron circuit discharges, the memristor resistance of expression membrane potential of neurons is reset, it is returned to high-impedance state, realize cell
The expression of film initial potential.
Further, regulation of the discharge cell to cynapse Array transfer efficiency, is according to neuron activity sequential phase
The plasticity STDP principles of pass are carried out respectively to each cynapse resistance:
After discharge cell sends electric signal, it is found that upper level neuron has also generated an action electricity after a short time
Position, the cynapse resistance being attached thereto becomes big, and transmission efficiency diminishes;After discharge cell sends electric signal, upper level nerve is found
An action potential has also been sent out before a bit of time by unit, and the cynapse resistance being attached thereto diminishes, and transmission efficiency becomes big.
Further, the reference voltage VRSize Criterion of Selecting is to ensure that when only a small number of input signals are input into, whole
Electric signal amplitude is less than reference voltage V after conjunctionR, and when compared with multiple input signals or synaptic efficacy higher, electric signal after integration
Amplitude is more than reference voltage VR
In the present invention, the memristor of cynapse array selects part volatibility bipolarity electric resistance changing device, expresses neuron
The memristor of the memristor of film potential selects volatibility electric resistance changing device, and MOS transistor T is from p-type transistor or other pressures
Control switch, anti-phase summation operation device, phase inverter, resistance, comparator, Spike signal generators etc. be ripe commercial device or
Equipment.Emulation completion is carried out by selected device and the neuron circuit for building, and with cynapse elementary cell.The neuron
Circuit can realize the integration discharging function in biological neuron, give expression to local hierarchical current potential, and cynapse has part volatibility,
The related plasticity of activity schedule can be expressed.
The present invention can realize the integration discharging function in biological neuron, give expression to local hierarchical current potential, to other god
The electric signal come through unit's transmission carries out space-time integration, meanwhile, the action potential of generation can with the electric signal of other neurons transmission
To be adjusted by related plasticity (Spike-timing-dependent plasticity, the STDP) rule of activity schedule
The plasticity of cynapse, realizes the time encoding to bursting activity.The cynapse has part volatibility, is more nearly biological true
Property.This neuron can provide elementary cell for hardware simulation cerebral nerve network structure, the god for overcoming prior art to exist
Postpone through first discharge time, it is difficult to realize the technical problems such as High Density Integration.This neuron circuit is used to construct class brain
Information processing system, can quickly process bulk information parallel, overcome computer of the tradition based on von Karman structure in image
Deficiency in terms of the intelligence such as identification, Self Adaptive Control, study, reasoning, decision-making.
Brief description of the drawings
Fig. 1 is neuron circuit proposed by the present invention;
Fig. 2 is the electrical property of memristor array of the present invention;
Fig. 3 is the electrical property of memristor of the present invention 2;
Fig. 4 is a kind of embodiment of neuron circuit;
Fig. 5 is the regulation of three kinds of situations and the related plasticity of activity schedule of neuron circuit emulation;
Fig. 6 is the neuron circuit based on part volatibility bipolarity electric resistance changing device proposed by the present invention.
Specific embodiment
Substantive distinguishing features of the invention are described further by example below in conjunction with the accompanying drawings.Explanation is needed herein
It is to be used to help understand the present invention for the explanation of these implementation methods, but does not constitute limitation of the invention.
Embodiment:
In this programme, cynapse array is memristor array, and adder constitutes anti-phase asking by operational amplifier combination resistance
And device, and then one phase inverter of connection will recover polarity of voltage, integrate the signal of input.The expression membrane potential of neurons
Memristor is memristor 2, is connected to inverter output.Neuron dendron, aixs cylinder are reached with wire table in the present invention.Memristor 2
The other end point two-way, meets divider resistance R all the wayc, another road connects comparator, and comparator output terminal connects Spike signal generators.Phase
Answering the selection of device will match with the resistance of memristor array and memristor 2, in the similar order of magnitude, be specifically shown in Fig. 4.
Cynapse array is memristor array, using Ni/Nb-SrTiO3/ Ti (strontium titanates/titanium of nickel/niobium doping) device, property
Can be such as Fig. 2.The simulation result of resistance variation characteristic and resistance retentivity when stimulating comprising the device pulse here.It is positive when applying
During scanning voltage, devices transition to low resistance state;When negative sense scanning voltage is applied, devices transition to high-impedance state;It is continuous when applying
Direct impulse signal when, the resistance of device is gradually reduced;When continuous negative-going pulse signal can be applied, the resistance of device by
It is cumulative to add;Meanwhile, the resistance state of device can slowly change over time, and one is returned to compared with high-impedance state by the spontaneous part of low-resistance.
In magnetron sputtering apparatus, with Nb-SrTiO3Monocrystalline is matrix, and Ti electrodes are plated in its bottom using magnetron sputtering, and top plates
Ni electrodes, prepare the half volatibility Ni/Nb-SrTiO with class synaptic function3/ Ti memory resistors.
The memristor of the expression membrane potential of neurons is memristor 2, and it uses Pt/WO3/ Pt (platinum/tungsten oxide/platinum) is complete
Full volatibility bipolarity electric resistance changing device, performance such as Fig. 3.When forward scan voltage is applied to 2V, device resistance can be dropped to
1M Ω, when further increasing forward scan voltage, device resistance can be reduced to smaller resistance;When apply negative sense scanning voltage to-
During 1.5V, device resistance is converted to 100M Ω, and device low resistance state keeps not living, and spontaneous can return to high-impedance state.The memristor
It is made as:In magnetron sputtering apparatus, to be covered with the monocrystalline silicon piece of certain thickness oxide layer as matrix, with Ti to stick
Layer, Pt is hearth electrode and top electrode, WO3It is functional layer, prepares the Pt/WO with complete volatibility3/ Pt memory resistors.
Emulated by neuron circuit, realize integration and discharging function of the neuron to electric signal.Fig. 4 is given
A kind of embodiment of neuron circuit, the cynapse array being made up of memristor array, adder, memristor 2 and electric discharge are single
Unit etc. is constituted.Memristor array is represented that memristor 2 is represented by MEMRISTOR-2 by MEMRISTOR-1.Other are conventional electricity
Sub- device.
Neuron circuit in Fig. 4 is so work:Memristor array is used as nerve synapse, Spike signals (these letters
Number come from the action potential of upper level neuron generation) via 3 cynapses, passed through after dendron enters anti-phase summation operation device
Phase inverter, carries out real-time integration, and change the resistance of memristor 2.Here, anti-phase sum operational amplifier is by the times magnification of signal
Number depends on resistance R2With the resistance value ratio of cynapse array, MOS transistor IRF450 because applying -5V signals, source electrode and drain electrode always
Between be conducting state.When the resistance of memristor 2 reduces to a certain extent, integrated signal is in resistance R6Partial pressure it is higher when, promote
Spike signal generators send the action potential as upper level neuron circuit below, point three road signals, all the way as whole
Close the action potential for sending that discharges and be delivered to next stage neuron circuit, controlling transistor switch IRF450, makes its source electrode all the way
Disconnected with drain electrode, another road feeds back to the cynapse at dendron, and is interacted with input signal, adjusts the transmission efficiency W of cynapsein
(action potential that the input signal of cynapse and the neuron are produced is respectively acting on memristor array two ends, according to 1 gust of memristor
Each memristor electric resistance changing mechanism in row, changes the resistance of device, and the application time difference of the two signals can influence the resistance of device
Value changes, reflect the related plasticity of activity schedule, i.e. STDP well).Memristor 2 is complete volatibility second-order memristor,
I.e. the device changes low resistance state after electric pulse effect, and the low resistance state keeps not living, and meeting is spontaneous over time to be returned to originally
High-impedance state, such as local hierarchical current potential in neuron.Here realize that the device of comparator effect is contained in the generation of Spike signals
Device, such as Fig. 4 dotted box portions.As resistance R6Partial pressure it is higher when, Spike signal generators can launch action potential, otherwise not
OK.
Be presented in Fig. 5 three kinds of neuron circuits to upper level neuron circuit action potential through cynapse array by adder
The example integrated is added in real time, and wherein PLUS represents anode in figure, and MINUS represents negative terminal:Set to work as and receive the prominent of signal
Touch more, such as 3, then neuronal cell membrane voltage can be more than threshold value, and electric discharge occurs, and (Fig. 5 a, wherein circuit indicate 3 cynapses
The artificial circuit of signal input, in the electric discharge figure of lower section three, first represents electric signal size after integration, second expression memristor
The change in resistance situation of device 2, the 3rd represents divider resistance (RC) voltage-to-ground situation of change);When the cynapse for receiving signal is less,
Such as 2, then not over threshold value, electric discharge does not occur, and (Fig. 5 b, wherein circuit indicate 2 cynapses to neuronal cell membrane voltage
The artificial circuit of signal input, in the electric discharge figure of lower section three, first represents electric signal size after integration, second expression memristor
The change in resistance situation of device 2, the 3rd represents divider resistance (RC) voltage-to-ground situation of change);When the cynapse for receiving signal is less,
Such as 2, but synaptic connection strengths or transmission efficiency are higher, then neuronal cell membrane voltage can exceed threshold value, and electric discharge (figure occurs
5c, wherein circuit indicate 2 artificial circuits of cynapse signal input, in the electric discharge figure of lower section three, after first represents integration
Electric signal size, second expression change in resistance situation of memristor 2, the 3rd represents divider resistance (RC) voltage-to-ground change feelings
Condition).There is the influence of the bonding strength to cynapse after action potential in neuron that Fig. 5 d are then provided, i.e., what activity schedule was related can
Plasticity.Three electric pulses of Fig. 5 d left figures are respectively expression upper level neuron circuit action potential, neuron circuit electric discharge
Action potential, both be superimposed because there is time difference after the actually applied memristor two ends in cynapse array voltage, Fig. 5 d
Right figure is the action potential of upper level neuron circuit action potential and neuron circuit electric discharge in the different time differences pair
The influence of the transmission efficiency of memristor in cynapse array.
Fig. 6 is the neuron circuit based on part volatibility bipolarity electric resistance changing device proposed by the present invention.With Fig. 1 not
Together, the memristor 2 of the circuit is turned using part volatibility bipolarity electric resistance changing device or complete non-volatile bipolarity resistance
When becoming device, because device resistance spontaneous can not return to high-impedance state, so in integrated signal VoutIncrease and memristor 2 between
Reset signal, is used to for device resistance to be reset to high-impedance state when neuron circuit discharges, and completes the nerve of the function of neuron
First circuit.
The present embodiment combines actual memory resistor resistance switching performance, is emulated by the neuron circuit for designing,
Demonstrating each essential characteristic of neuron circuit design includes number, the neuronal transmission efficiency of cynapse to neuronal integration
Influence of electric discharge etc. is successfully realized.Illustrate that neuron circuit design is rational, can be by actual physics device
What part was realized.
The above is presently preferred embodiments of the present invention, but the present invention should not be limited to the embodiment and accompanying drawing institute
Disclosure.So every do not depart from the lower equivalent or modification for completing of spirit disclosed in this invention, guarantor of the present invention is both fallen within
The scope of shield.
Claims (7)
1. a kind of neuron circuit based on memory resistor, it is characterised in that including cynapse array, switching tube (T), adder
(OP1) memristor, comparator (OP2), the Spike signal generators of membrane potential of neurons, are expressed;Wherein:
The cynapse array is used to receive the action potential that upper level neuron circuit is transmitted, and it includes several parts arranged side by side
Volatibility memristor, each memristor one end is connected with each aixs cylinder of upper level neuron circuit, and other end tandem is warp all the way
Switch (T) is connected with adder (OP1) input;
The adder (OP1) is used to integrate the action potential of each input of cynapse array, for adjusting expression neuron membrane electricity
The resistance of the memristor of position, so as to realize the simulation to membrane potential of neurons;
The memristor of the expression membrane potential of neurons is complete volatibility memristor, for the part of imictron cell membrane
Graded potential;One terminates the output end of the adder (OP1), and the other end is divided into two-way, divider resistance (R is met all the wayC) connect
Ground;Another road connects comparator (OP2), comparator (OP2) input will be delivered to after action potential partial pressure after the integration of acquisition;
Another input termination reference voltage V of comparator (OP2)R, for comparing divider resistance (RC) voltage-to-ground and reference electricity
Pressure VRSize;As divider resistance (RC) voltage-to-ground be more than VRWhen, conduction level is exported, otherwise export cut-off level;
The Spike signal generators input is connected with comparator (OP2) output end, and Spike signal generators export three tunnels
Signal, all the way the control pole with the switching tube (T) be connected, for the action of controlling switch pipe (T);Second road signal connects cynapse
The tandem end of array, the transmission efficiency for adjusting cynapse array;3rd road signal connects next stage neuron circuit, used as next
Level neuron circuit input signal.Usually during comparator (OP2) output cut-off level, Spike signal generators are stopped, and make
Switching tube (T) is turned on;When comparator (OP2) exports conduction level, Spike signal generators export a cut-off signals, make
Switching tube (T) disconnects.
2. neuron circuit according to claim 1, it is characterised in that during work, via cynapse array be input into it is upper
One-level neuron signal is added by adder (OP1), enters horizontal electrical signal real-time integration, and electric signal amplitude rises to after integration
During threshold value, reduce the resistance of the memristor of expression membrane potential of neurons so that divider resistance (RC) voltage-to-ground increase;When point
Piezoresistance (RC) voltage-to-ground exceed reference voltage VRWhen, comparator (OP2) output conduction level notifies Spike signal generators
According to the electric signal electric discharge that the simulation for prestoring is biological, sending for action potential is realized, complete the integration electric discharge work(of neuron
Energy;The reference voltage VRBe according to expression membrane potential of neurons memristor resistance and divider resistance (RC) resistance size
And the amplitude of neuron action potential mutually weigh depending on set a constant pressure.
3. neuron circuit according to claim 1 and 2, it is characterised in that each cynapse in the cynapse array, adopts
Realized with part volatibility memristor.
4. neuron circuit according to claim 1 and 2, it is characterised in that the computing that the adder (OP1) uses is put
Big device.
5. neuron circuit according to claim 1 and 2, it is characterised in that the memristor of the expression membrane potential of neurons
Device selects part volatibility device or complete non-volatile device, and the output end all the way in Spike signal generators is refreshing with expression
Memristor through first film potential is connected with the output end of adder (OP1), for when neuron circuit discharges, resetting expression god
Through the memristor resistance of first film potential, it is returned to high-impedance state, realize the expression of cell membrane initial potential.
6. neuron circuit according to claim 1 and 2, it is characterised in that the Spike signal generators are to cynapse battle array
The regulation of the defeated efficiency of biographies, is that each cynapse resistance is entered respectively according to the related plasticity STDP principles of neuron activity sequential
OK:After Spike signal generators send electric signal, it is found that upper level neuron has also sent out an action after a short time
Current potential, the cynapse resistance being attached thereto becomes big, and transmission efficiency diminishes;After discharge cell sends electric signal, upper level god is found
An action potential is also sent out before a bit of time through unit, the cynapse resistance being attached thereto diminishes, transmission efficiency becomes big.
7. neuron circuit according to claim 1 and 2, it is characterised in that the reference voltage VRSize Criterion of Selecting is
Ensure that electric signal amplitude is less than reference voltage V after integration when only a small number of input signals are input intoR, and when compared with multiple input signals
Or synaptic efficacy it is higher when, after integration electric signal amplitude be more than reference voltage VR。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611235356.9A CN106845634B (en) | 2016-12-28 | 2016-12-28 | A kind of neuron circuit based on memory resistor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611235356.9A CN106845634B (en) | 2016-12-28 | 2016-12-28 | A kind of neuron circuit based on memory resistor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106845634A true CN106845634A (en) | 2017-06-13 |
CN106845634B CN106845634B (en) | 2018-12-14 |
Family
ID=59113084
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611235356.9A Active CN106845634B (en) | 2016-12-28 | 2016-12-28 | A kind of neuron circuit based on memory resistor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106845634B (en) |
Cited By (34)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107742153A (en) * | 2017-10-20 | 2018-02-27 | 华中科技大学 | A kind of neuron circuit with stable state plasticity based on memristor |
CN107909146A (en) * | 2017-11-13 | 2018-04-13 | 中国科学院微电子研究所 | Neuron circuit based on volatibility threshold transitions device |
CN108664735A (en) * | 2018-05-11 | 2018-10-16 | 华中科技大学 | The implementation method of STDP pulse design methods and diversification STDP based on multivalue memristor |
CN109146073A (en) * | 2017-06-16 | 2019-01-04 | 华为技术有限公司 | A kind of neural network training method and device |
CN109165731A (en) * | 2018-08-09 | 2019-01-08 | 清华大学 | Electronic neuron network and its parameter setting method |
CN109255437A (en) * | 2018-08-17 | 2019-01-22 | 郑州轻工业学院 | A kind of memristor nerve network circuit of flexibly configurable |
CN109449289A (en) * | 2018-11-01 | 2019-03-08 | 中国科学院宁波材料技术与工程研究所 | A kind of bionical memristor of the nerve synapse of light stimulus and preparation method thereof |
CN109447250A (en) * | 2018-09-14 | 2019-03-08 | 华中科技大学 | A kind of artificial neuron based on battery effect in memristor |
CN109978019A (en) * | 2019-03-07 | 2019-07-05 | 东北师范大学 | Image steganalysis simulation mixes memristor equipment and preparation with number, realizes STDP learning rules and image steganalysis method |
CN109977470A (en) * | 2019-02-20 | 2019-07-05 | 华中科技大学 | A kind of circuit and its operating method based on memristor Hopfield neural fusion sparse coding |
CN110059816A (en) * | 2019-04-09 | 2019-07-26 | 南京邮电大学 | A kind of neural network element circuit based on memristor |
CN110111234A (en) * | 2019-04-11 | 2019-08-09 | 上海集成电路研发中心有限公司 | A kind of image processing system framework neural network based |
CN110163365A (en) * | 2019-05-29 | 2019-08-23 | 北京科易达知识产权服务有限公司 | A kind of spiking neuron circuit applied to memristor synapse array |
CN110163364A (en) * | 2019-04-28 | 2019-08-23 | 南京邮电大学 | A kind of neural network element circuit based on memristor bridge cynapse |
CN110309908A (en) * | 2019-06-11 | 2019-10-08 | 北京大学 | FeFET-CMOS mixed pulses neuron based on ferroelectric transistor |
CN110443356A (en) * | 2019-08-07 | 2019-11-12 | 南京邮电大学 | A kind of current mode neural network based on more resistance state memristors |
CN110751273A (en) * | 2018-07-22 | 2020-02-04 | 徐志强 | Neuron and synapse simulation assembly |
CN110837253A (en) * | 2019-10-31 | 2020-02-25 | 华中科技大学 | Intelligent addressing system based on memristor synapse |
CN110991610A (en) * | 2019-11-28 | 2020-04-10 | 华中科技大学 | Probabilistic neuron circuit, probabilistic neural network topological structure and application thereof |
CN111129297A (en) * | 2019-12-30 | 2020-05-08 | 北京大学 | Method and system for realizing diversity STDP of memristive synapse device |
CN111275177A (en) * | 2020-01-16 | 2020-06-12 | 北京大学 | Full memristor neural network and preparation method and application thereof |
CN111401540A (en) * | 2020-03-09 | 2020-07-10 | 北京航空航天大学 | Neuron model construction method and neuron model |
CN111771214A (en) * | 2018-01-31 | 2020-10-13 | 阿姆有限公司 | Correlated electronic switching element for brain-based computing |
CN111967589A (en) * | 2020-08-21 | 2020-11-20 | 清华大学 | Neuron analog circuit, driving method thereof and neural network device |
CN111958599A (en) * | 2020-08-17 | 2020-11-20 | 湖南大学 | Self-repairing control system based on astrocytes and intelligent robot arm |
CN112053726A (en) * | 2020-09-09 | 2020-12-08 | 哈尔滨工业大学 | Flash memory mistaken erasure data recovery method based on Er-state threshold voltage distribution |
CN112598124A (en) * | 2020-12-28 | 2021-04-02 | 清华大学 | Neuron analog circuit and neural network device |
CN112906880A (en) * | 2021-04-08 | 2021-06-04 | 华中科技大学 | Adaptive neuron circuit based on memristor |
CN113191492A (en) * | 2021-04-14 | 2021-07-30 | 华中科技大学 | Synapse training architecture |
CN113748433A (en) * | 2019-04-25 | 2021-12-03 | Hrl实验室有限责任公司 | Active memristor-based spiking neuromorphic circuit for motion detection |
CN113807161A (en) * | 2020-06-11 | 2021-12-17 | 华邦电子股份有限公司 | Writing method of spike timing dependent plasticity and synapse array device |
CN114239466A (en) * | 2021-12-22 | 2022-03-25 | 华中科技大学 | Circuit for realizing multi-mode information fusion association based on memristor BAM and application thereof |
CN115688897A (en) * | 2023-01-03 | 2023-02-03 | 浙江大学杭州国际科创中心 | Low-power-consumption compact Relu activation function neuron circuit |
CN116663632A (en) * | 2023-08-02 | 2023-08-29 | 华中科技大学 | Intelligent sensing system integrating sensing, storage and calculation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012006471A1 (en) * | 2010-07-07 | 2012-01-12 | Qualcomm Incorporated | Methods and systems for memristor-based neuron circuits |
WO2013044143A1 (en) * | 2011-09-21 | 2013-03-28 | Qualcomm Incorporated | Method and apparatus for structural delay plasticity in spiking neural networks |
CN103941581A (en) * | 2014-04-17 | 2014-07-23 | 广西大学 | Single-neuron PID controller based on memory resistors |
CN105701541A (en) * | 2016-01-13 | 2016-06-22 | 哈尔滨工业大学深圳研究生院 | Circuit structure based on memristor pulse nerve network |
-
2016
- 2016-12-28 CN CN201611235356.9A patent/CN106845634B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012006471A1 (en) * | 2010-07-07 | 2012-01-12 | Qualcomm Incorporated | Methods and systems for memristor-based neuron circuits |
WO2013044143A1 (en) * | 2011-09-21 | 2013-03-28 | Qualcomm Incorporated | Method and apparatus for structural delay plasticity in spiking neural networks |
CN103941581A (en) * | 2014-04-17 | 2014-07-23 | 广西大学 | Single-neuron PID controller based on memory resistors |
CN105701541A (en) * | 2016-01-13 | 2016-06-22 | 哈尔滨工业大学深圳研究生院 | Circuit structure based on memristor pulse nerve network |
Non-Patent Citations (3)
Title |
---|
MYONGLAE CHU: "Neuromorphic Hardware System for Visual", 《IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS》 * |
张小红: "新型忆阻细胞神经网络的建模及电路仿真", 《系统仿真学报》 * |
李志军: "改进型细胞神经网络实现的忆阻器混沌电路", 《物理学报》 * |
Cited By (57)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109146073A (en) * | 2017-06-16 | 2019-01-04 | 华为技术有限公司 | A kind of neural network training method and device |
CN109146073B (en) * | 2017-06-16 | 2022-05-24 | 华为技术有限公司 | Neural network training method and device |
US11475300B2 (en) | 2017-06-16 | 2022-10-18 | Huawei Technologies Co., Ltd. | Neural network training method and apparatus |
CN107742153A (en) * | 2017-10-20 | 2018-02-27 | 华中科技大学 | A kind of neuron circuit with stable state plasticity based on memristor |
CN107742153B (en) * | 2017-10-20 | 2020-02-21 | 华中科技大学 | Memristor-based neuron circuit with steady-state plasticity |
CN107909146A (en) * | 2017-11-13 | 2018-04-13 | 中国科学院微电子研究所 | Neuron circuit based on volatibility threshold transitions device |
CN111771214A (en) * | 2018-01-31 | 2020-10-13 | 阿姆有限公司 | Correlated electronic switching element for brain-based computing |
CN108664735A (en) * | 2018-05-11 | 2018-10-16 | 华中科技大学 | The implementation method of STDP pulse design methods and diversification STDP based on multivalue memristor |
CN108664735B (en) * | 2018-05-11 | 2020-06-09 | 华中科技大学 | STDP pulse design method based on multivalued memristor and realization method of diversified STDP |
CN110751273A (en) * | 2018-07-22 | 2020-02-04 | 徐志强 | Neuron and synapse simulation assembly |
CN109165731A (en) * | 2018-08-09 | 2019-01-08 | 清华大学 | Electronic neuron network and its parameter setting method |
CN109165731B (en) * | 2018-08-09 | 2020-06-30 | 清华大学 | Electronic neural network and parameter setting method thereof |
CN109255437A (en) * | 2018-08-17 | 2019-01-22 | 郑州轻工业学院 | A kind of memristor nerve network circuit of flexibly configurable |
CN109255437B (en) * | 2018-08-17 | 2019-06-14 | 郑州轻工业学院 | A kind of memristor nerve network circuit of flexibly configurable |
CN109447250B (en) * | 2018-09-14 | 2020-07-10 | 华中科技大学 | Artificial neuron based on battery effect in memristor |
CN109447250A (en) * | 2018-09-14 | 2019-03-08 | 华中科技大学 | A kind of artificial neuron based on battery effect in memristor |
CN109449289A (en) * | 2018-11-01 | 2019-03-08 | 中国科学院宁波材料技术与工程研究所 | A kind of bionical memristor of the nerve synapse of light stimulus and preparation method thereof |
CN109977470A (en) * | 2019-02-20 | 2019-07-05 | 华中科技大学 | A kind of circuit and its operating method based on memristor Hopfield neural fusion sparse coding |
CN109977470B (en) * | 2019-02-20 | 2020-10-30 | 华中科技大学 | Circuit for sparse coding of memristive Hopfield neural network and operation method thereof |
CN109978019A (en) * | 2019-03-07 | 2019-07-05 | 东北师范大学 | Image steganalysis simulation mixes memristor equipment and preparation with number, realizes STDP learning rules and image steganalysis method |
CN110059816B (en) * | 2019-04-09 | 2022-08-16 | 南京邮电大学 | Memristor-based neural network unit circuit |
CN110059816A (en) * | 2019-04-09 | 2019-07-26 | 南京邮电大学 | A kind of neural network element circuit based on memristor |
CN110111234B (en) * | 2019-04-11 | 2023-12-15 | 上海集成电路研发中心有限公司 | Image processing system architecture based on neural network |
CN110111234A (en) * | 2019-04-11 | 2019-08-09 | 上海集成电路研发中心有限公司 | A kind of image processing system framework neural network based |
CN113748433A (en) * | 2019-04-25 | 2021-12-03 | Hrl实验室有限责任公司 | Active memristor-based spiking neuromorphic circuit for motion detection |
CN113748433B (en) * | 2019-04-25 | 2023-03-28 | Hrl实验室有限责任公司 | Active memristor-based spiking neuromorphic circuit for motion detection |
CN110163364A (en) * | 2019-04-28 | 2019-08-23 | 南京邮电大学 | A kind of neural network element circuit based on memristor bridge cynapse |
CN110163364B (en) * | 2019-04-28 | 2022-08-30 | 南京邮电大学 | Neural network unit circuit based on memristor bridge synapse |
CN110163365A (en) * | 2019-05-29 | 2019-08-23 | 北京科易达知识产权服务有限公司 | A kind of spiking neuron circuit applied to memristor synapse array |
CN110309908A (en) * | 2019-06-11 | 2019-10-08 | 北京大学 | FeFET-CMOS mixed pulses neuron based on ferroelectric transistor |
CN110443356A (en) * | 2019-08-07 | 2019-11-12 | 南京邮电大学 | A kind of current mode neural network based on more resistance state memristors |
CN110443356B (en) * | 2019-08-07 | 2022-03-25 | 南京邮电大学 | Current type neural network based on multi-resistance state memristor |
CN110837253A (en) * | 2019-10-31 | 2020-02-25 | 华中科技大学 | Intelligent addressing system based on memristor synapse |
CN110991610A (en) * | 2019-11-28 | 2020-04-10 | 华中科技大学 | Probabilistic neuron circuit, probabilistic neural network topological structure and application thereof |
CN110991610B (en) * | 2019-11-28 | 2022-08-05 | 华中科技大学 | Probability determination method for nondeterministic problem |
CN111129297B (en) * | 2019-12-30 | 2022-04-19 | 北京大学 | Method and system for realizing diversity STDP of memristive synapse device |
CN111129297A (en) * | 2019-12-30 | 2020-05-08 | 北京大学 | Method and system for realizing diversity STDP of memristive synapse device |
CN111275177B (en) * | 2020-01-16 | 2022-10-21 | 北京大学 | Full memristor neural network and preparation method and application thereof |
CN111275177A (en) * | 2020-01-16 | 2020-06-12 | 北京大学 | Full memristor neural network and preparation method and application thereof |
CN111401540A (en) * | 2020-03-09 | 2020-07-10 | 北京航空航天大学 | Neuron model construction method and neuron model |
CN113807161A (en) * | 2020-06-11 | 2021-12-17 | 华邦电子股份有限公司 | Writing method of spike timing dependent plasticity and synapse array device |
CN111958599A (en) * | 2020-08-17 | 2020-11-20 | 湖南大学 | Self-repairing control system based on astrocytes and intelligent robot arm |
CN111967589B (en) * | 2020-08-21 | 2023-12-26 | 清华大学 | Neuron simulation circuit, driving method thereof and neural network device |
CN111967589A (en) * | 2020-08-21 | 2020-11-20 | 清华大学 | Neuron analog circuit, driving method thereof and neural network device |
CN112053726A (en) * | 2020-09-09 | 2020-12-08 | 哈尔滨工业大学 | Flash memory mistaken erasure data recovery method based on Er-state threshold voltage distribution |
CN112598124A (en) * | 2020-12-28 | 2021-04-02 | 清华大学 | Neuron analog circuit and neural network device |
CN112598124B (en) * | 2020-12-28 | 2022-12-20 | 清华大学 | Neuron analog circuit and neural network device |
CN112906880B (en) * | 2021-04-08 | 2022-04-26 | 华中科技大学 | Adaptive neuron circuit based on memristor |
CN112906880A (en) * | 2021-04-08 | 2021-06-04 | 华中科技大学 | Adaptive neuron circuit based on memristor |
CN113191492B (en) * | 2021-04-14 | 2022-09-27 | 华中科技大学 | Synapse training device |
CN113191492A (en) * | 2021-04-14 | 2021-07-30 | 华中科技大学 | Synapse training architecture |
CN114239466A (en) * | 2021-12-22 | 2022-03-25 | 华中科技大学 | Circuit for realizing multi-mode information fusion association based on memristor BAM and application thereof |
CN114239466B (en) * | 2021-12-22 | 2024-06-04 | 华中科技大学 | Circuit for realizing multi-mode information fusion association based on memristor BAM and application thereof |
CN115688897A (en) * | 2023-01-03 | 2023-02-03 | 浙江大学杭州国际科创中心 | Low-power-consumption compact Relu activation function neuron circuit |
CN115688897B (en) * | 2023-01-03 | 2023-03-31 | 浙江大学杭州国际科创中心 | Low-power-consumption compact Relu activation function neuron circuit |
CN116663632A (en) * | 2023-08-02 | 2023-08-29 | 华中科技大学 | Intelligent sensing system integrating sensing, storage and calculation |
CN116663632B (en) * | 2023-08-02 | 2023-10-10 | 华中科技大学 | Intelligent sensing system integrating sensing, storage and calculation |
Also Published As
Publication number | Publication date |
---|---|
CN106845634B (en) | 2018-12-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106845634B (en) | A kind of neuron circuit based on memory resistor | |
CN109447250B (en) | Artificial neuron based on battery effect in memristor | |
US10650308B2 (en) | Electronic neuromorphic system, synaptic circuit with resistive switching memory and method of performing spike-timing dependent plasticity | |
CN103078054B (en) | Unit, device and method for simulating biological neuron and neuronal synapsis | |
TWI509537B (en) | Electronic learning synapse with spike-timing dependent plasticity using memory-switching elements | |
Wu et al. | Homogeneous spiking neuromorphic system for real-world pattern recognition | |
CN106779059A (en) | A kind of Circuit of Artificial Neural Networks of the Pavlov associative memory based on memristor | |
CN102610274B (en) | Weight adjustment circuit for variable-resistance synapses | |
Brivio et al. | Extended memory lifetime in spiking neural networks employing memristive synapses with nonlinear conductance dynamics | |
CN103778468A (en) | RRAM-based new type neural network circuit | |
US10740672B2 (en) | Capacitative artificial neural networks | |
KR20150034900A (en) | Synapse circuit for connecting neuron circuits, unit cell composing neuromorphic circuit, and neuromorphic circuit | |
CN103580668A (en) | Associative memory circuit based on memory resistor | |
CN103078055A (en) | Unit, device and method for simulating biological neuronal synapsis | |
CN107122828B (en) | Circuit structure, driving method thereof and neural network | |
CN109787592B (en) | Random nerve pulse generator | |
CN110232440A (en) | Spiking neuron circuit based on ferroelectric transistor | |
Erokhin et al. | Polymeric elements for adaptive networks | |
Kuncic et al. | Neuromorphic information processing with nanowire networks | |
Mahalanabis et al. | Demonstration of spike timing dependent plasticity in CBRAM devices with silicon neurons | |
JP2021033415A (en) | Spiking neural network device and training method thereof | |
Milo et al. | Resistive switching synapses for unsupervised learning in feed-forward and recurrent neural networks | |
CN112906880B (en) | Adaptive neuron circuit based on memristor | |
CN109165731B (en) | Electronic neural network and parameter setting method thereof | |
CN105373829B (en) | A kind of full Connection Neural Network structure |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |