CN108777153A - A kind of its plasticity modulator approach of multiterminal input cynapse device - Google Patents
A kind of its plasticity modulator approach of multiterminal input cynapse device Download PDFInfo
- Publication number
- CN108777153A CN108777153A CN201810516810.0A CN201810516810A CN108777153A CN 108777153 A CN108777153 A CN 108777153A CN 201810516810 A CN201810516810 A CN 201810516810A CN 108777153 A CN108777153 A CN 108777153A
- Authority
- CN
- China
- Prior art keywords
- signal
- multiterminal
- artificial synapse
- electric conductivity
- conductivity value
- 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
- 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
- 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
- G11C11/21—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements
- G11C11/22—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using ferroelectric elements
-
- 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
- G11C11/21—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements
- G11C11/34—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices
- G11C11/40—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors
- G11C11/401—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors forming cells needing refreshing or charge regeneration, i.e. dynamic cells
- G11C11/402—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors forming cells needing refreshing or charge regeneration, i.e. dynamic cells with charge regeneration individual to each memory cell, i.e. internal refresh
- G11C11/4023—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors forming cells needing refreshing or charge regeneration, i.e. dynamic cells with charge regeneration individual to each memory cell, i.e. internal refresh using field effect transistors
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Neurology (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Non-Volatile Memory (AREA)
Abstract
The invention discloses a kind of multiterminal input cynapse device and its plasticity modulator approach, device is electric conductivity value non-volatile device, and device electric conductivity value changes in the case where external drive continuously regulates and controls, and device includes:The front port for making its electric conductivity value become larger after receiving signal receives presynaptic neuron and transmits signal for connecting presynaptic neuron;The rear port for making its electric conductivity value become smaller after receiving signal receives postsynaptic neuron and transmits signal for connecting postsynaptic neuron;Make that front port input signal changes to electric conductivity value control law or rear port input signal is to the changed control terminal of electric conductivity value control law after receiving signal, for receiving control signal;And the reading port that its electric conductivity value remains unchanged after receiving the signal less than threshold value, for the input signal output signal output according to the input signal of front port, the input signal of rear port and control terminal.The artificial synapse device realizes that study reads parallel carry out with learning outcome.
Description
Technical field
The invention belongs to artificial neural network technology fields, and more particularly, to a kind of multiterminal input cynapse device, it can
Plasticity modulator approach.
Background technology
Under the big data epoch, mass data large-scale calculations storage and transmission the problem of it is urgently to be resolved hurrily.Traditional von Neumann
Computer architecture is detached since data are calculated with storage, is connected with bus, keeps the transmission speed of data significantly limited.And in human brain
It calculates the information processing manner calculated with memory parallel and provides a kind of new approaches, artificial neural network to break through von Neumann bottleneck
The research of network becomes the effective and feasible scheme for calculating and being carried out with memory parallel.
Human brain nervous system is huge and complicated, about by 1011A neuron and 1014A nerve synapse is constituted.Human brain simultaneously
When row memory is with received information is calculated, each neuron and cynapse are synchronously storing and are handling information, work as outside stimulus
When being transferred to cynapse by the variation of neuropotential by nervous system, synapse weight can be according to the type and intensity of stimulation correspondingly
Enhancing occurs and weakens, such cynapse behavior is referred to as synaptic plasticity.Synaptic plasticity is that human brain transmits and handle information
When neural biomolecular science basis and most basic cognitive activities --- the basis of learning and memory of human brain.In nerveous system
In system, most common synaptic plasticity includes:Iuntercellular long term potentiation (LTP, Long Term Potentiation), it is long when
Journey inhibits (LTD, Long Term Depression), pulse sequence to rely on synaptic plasticity (STDP, Spike Timing
Dependent Plasticity), pulse frequency rely on synaptic plasticity (SRDP, Spike Rate Dependent
Plasticity) etc..Pulse sequence relies on synaptic plasticity STDP and is used as one kind in insect or even human brain difference neural circuitry
The cynapse behavior being found has become the neural network learning rule for being widely recognized as receiving.
STDP learning rules and the sequencing of presynaptic neuron activity and postsynaptic neuron activity in time are tight
The weight of close correlation, cynapse changes as presynaptic and postsynaptic neuron pulse reach the time difference decision of the cynapse.If prominent
Neuron pulse ratio postsynaptic neuron pulse reaches the cynapse earlier before touching, then the weight enhancing of cynapse, between two neurons
Relation intensity also enhance, and enhance degree with two pulses interval time increase and reduce;If presynaptic neuron arteries and veins
Punching arrives later the cynapse than postsynaptic neuron pulse, then the weight of cynapse weakens, the relation intensity between two neurons
The degree for weakening, and weakening increases also with the interval time of two pulses and is reduced.The variable quantity of synapse weight and front and back pulse
The time difference of arrival meets exponential relationship.Fig. 1 is the curve that specific STDP pulse sequences rely on synaptic plasticity learning rule
Schematic diagram.
Currently, using single artificial synapse device realize STDP functions research in, used device is mostly that resistive is deposited
The two-terminal devices such as reservoir, phase transition storage or ferromagnetic store, therefore, not only the modification scope of single cynapse device and regulation and control are smart
Degree is restricted, and since only there are two input ports for device, the learning manipulation of artificial synapse device and the read operation of weight must
It must carry out step by step, complicate the learning process of artificial synapse device.In addition, in human brain neural network between neuron and cynapse
Connection is complicated, and two-terminal device constrains single cynapse device being integrally interconnected in neural network.Therefore, be dedicated to three ends or even
The simulated implementation of various synaptic plasticities is that artificial neural network is important grinds for development in the artificial synapse device of more multiterminal regulation and control
Study carefully one of direction.
Invention content
For the disadvantages described above or Improvement requirement of the prior art, the present invention provides a kind of its plasticity of multiterminal cynapse device
Modulator approach, it is intended to which solve existing cynapse device causes study and learning outcome reading that cannot carry out parallel using two-port
Technical problem.
To achieve the above object, the present invention provides a kind of multiterminal cynapse device, and the multiterminal artificial synapse device is conductance
It is worth non-volatile device, device electric conductivity value changes in the case where external drive continuously regulates and controls, and device includes:
The front port I for making its electric conductivity value become larger after receiving signalpre, for connecting presynaptic neuron, receive the presynaptic
Neuronal transmission signal;
The rear port I for making its electric conductivity value become smaller after receiving signalpost, for connecting postsynaptic neuron, receive cynapse
Neuronal transmission signal afterwards;
Make front port I after receiving signalpreInput signal changes to electric conductivity value control law or rear port Ipost
Input signal control terminal I changed to electric conductivity value control lawc, for receiving control signal;And
The reading port I that its electric conductivity value remains unchanged after receiving the signal less than threshold valueread, for according to front port
IpreInput signal, rear port IpostInput signal and control terminal IcInput signal output signal output, output signal is same
The electric conductivity value of multiterminal artificial synapse device is related.
Preferably, the multiterminal artificial synapse device includes multiple control terminal Ic。
Preferably, the multiterminal artificial synapse device is two dimensional surface device or three-dimensional structure device.
Preferably, 3 D stereo multiterminal entering apparatus is bipolar transistor devices or FET device.
Preferably, FET device is NOR Flash devices, NAND Flash devices, SONOS devices, floating boom
Transistor device, ferro-electric field effect transistor device, spin fet device or oxide field-effect transistor device.
Preferably, when multiterminal artificial synapse device is floating gate transistor device or SONOS devices, grid is as multiterminal
The rear port I of artificial synapse devicepost, front port I of the body electrode as multiterminal artificial synapse devicepre, drain electrode is as multiterminal
The reading port I of artificial synapse deviceread, control terminal I of the source electrode end as multiterminal artificial synapse devicec;
Wherein, floating transistor includes the gate oxide for stacking gradually arrangement, floating gate layer, floating boom insulating layer and substrate layer;
It is equipped with grid above gate oxide, drain electrode and source electrode are equipped with successively in substrate layer both sides, body electrode is equipped in substrate layer;
SONOS devices include the upper oxidation insulating layer for stacking gradually placement, silicon nitride layer, lower oxidation insulating layer and substrate
Layer;It is equipped with grid above upper oxidation insulating layer, is provided at both ends with drain electrode and source electrode in substrate layer, body electrode connects substrate
Layer.
Preferably, when multiterminal artificial synapse device is ferro-electric field effect transistor, grid is as multiterminal artificial synapse device
The front port I of partpre, rear port I of the body electrode as multiterminal artificial synapse devicepost, drain electrode is as multiterminal artificial synapse device
The reading port I of partread, control terminal I of the source electrode end as multiterminal artificial synapse devicec;
Ferro-electric field effect transistor includes the ferroelectric material layer, insulating layer and substrate layer for stacking gradually arrangement, in ferroelectricity material
It is provided with grid above the bed of material, is provided at both ends with drain electrode and source electrode in substrate layer, body electrode connects substrate layer.
As another aspect of the present invention, the present invention provides the multistage cynapse based on above-mentioned multiterminal artificial synapse device can
Plasticity modulator approach, includes the following steps:
In front port IpreInput pulse signal sequence, by rear port IpostIt is grounded, each pulse letter in sequences of pulsed signals
Number the electric conductivity value of artificial synapse device can be made to increase, and in sequences of pulsed signals the amplitude of the latter pulse signal be more than it is previous
The amplitude of pulse signal or the width of the latter pulse signal are more than the width of previous pulse signal, make artificial synapse device
The increased degree of electric conductivity value of part becomes larger, and realizes enhanced synaptic plasticity function;Or
In rear port IpostInput pulse signal sequence, by front port IpreIt is grounded, each pulse letter in sequences of pulsed signals
Number the electric conductivity value of artificial synapse device can be made to reduce, and in sequences of pulsed signals the amplitude of the latter pulse signal be more than it is previous
The amplitude of pulse signal or the width of the latter pulse signal are more than the width of previous pulse signal, make artificial synapse device
The degree that the electric conductivity value of part reduces becomes larger, and realizes suppressive synaptic plasticity function.
As another aspect of the present invention, the present invention provide the pulse sequence based on above-mentioned multiterminal artificial synapse device according to
Rely synaptic plasticity modulator approach, includes the following steps:
Front port IpreInput presynaptic pulse signal Ppre, rear port IpostInput postsynaptic pulse signal Ppost;Presynaptic
The time difference Δ t of pulse signal and postsynaptic pulse signal is more than zero, and effective amplitude of superposed signal is more than zero, and superposed signal
Effective amplitude poor Δ t increases and reduces at any time;Realize that enhanced pulse sequence relies on synaptic plasticity function;Or
Front port IpreInput presynaptic pulse signal Ppre, rear port IpostInput postsynaptic pulse signal Ppost;Presynaptic
The time difference Δ t of pulse signal and postsynaptic pulse signal is less than zero, and effective amplitude of superposed signal is less than zero, and superposed signal
Effective amplitude poor Δ t increases and reduces at any time, realize suppressive pulse sequence dependence synaptic plasticity function.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
1, multiterminal artificial synapse device provided by the invention is equipped with front port Ipre, rear port Ipost, read port IreadAnd
Control port Ic.Wherein, front port Ipre, rear port IpostIt is respectively used to receive the signal and postsynaptic neuronal of presynaptic neuron
The signal of member, under the excitation of above-mentioned input signal, the electric conductivity value of device changes, and via reading port IreadOr it reads
Exit port IreadWith control port IcElectric conductivity value is read after cooperation, realizes that the multiterminal artificial synapse device receives signal and believes with output
It is number parallel.
2, the modulation strategy of multiterminal artificial synapse device provided by the invention, may be implemented it is multistage regulation and control and pulse sequence according to
Rely synaptic plasticity function, has widened the modification scope and degree of regulation of artificial synapse device synaptic plasticity, realized study
Parallel progress is read with learning outcome, feasibility is provided applied to the integrated of artificial neural network for single artificial synapse device
With diversity.
Description of the drawings
Fig. 1 is the structural schematic diagram of multiterminal artificial synapse device provided by the invention;
Fig. 2 is the curve synoptic diagram that the present invention is based on that STDP pulse sequences rely on synaptic plasticity learning rule;
Fig. 3 is the structural schematic diagram of the artificial synapse device provided by the invention based on floating transistor;
Fig. 4 is that the artificial synapse device provided by the invention based on floating transistor uses front port input signal rear port
Contact patch realizes the schematic diagram of multistage conductance regulation and control;
Fig. 5 is that the artificial synapse device provided by the invention based on floating transistor uses rear port input signal front port
Contact patch realizes the schematic diagram of multistage conductance regulation and control;
Fig. 6 is the multistage regulation and control result figure of conductance of the artificial synapse device provided by the invention based on floating transistor;
Fig. 7 is that there are the triangular waves of difference of injection time in input for the artificial synapse device provided by the invention based on floating transistor
The schematic diagram of enhanced synaptic plasticity function is realized under pulse;
Fig. 8 is that there are the triangular waves of difference of injection time in input for the artificial synapse device provided by the invention based on floating transistor
The schematic diagram of suppressive synaptic plasticity function is realized under pulse;
Fig. 9 is that there are the square wave arteries and veins of difference of injection time in input for the artificial synapse device provided by the invention based on floating transistor
Sweep away the schematic diagram for realizing enhanced synaptic plasticity function;
Figure 10 is that there are the square waves of difference of injection time in input for the artificial synapse device provided by the invention based on floating transistor
The schematic diagram of suppressive synaptic plasticity function is realized under pulse;
Figure 11 is that the structure of the artificial synapse device provided by the invention based on four end iron field effect transistor devices is shown
It is intended to;
Figure 12 is the structural schematic diagram of the artificial synapse device provided by the invention based on SONOS devices.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
The present invention provides a kind of multiterminal artificial synapse device and its plasticity modulator approaches, and the device is in the connection presynaptic
The front port I of neuronpreOr the rear port I of connection postsynaptic neuronpostContinuous impulse excitation under, have it is multistage
Conductance modulation characteristic is based on this characteristic, adjustment front port IpreWith rear port IpostIt the waveform of input signal and time difference, can be with
Simulated implementation pulse sequence dependent form synaptic plasticity.
As shown in Figure 1, multiterminal artificial synapse device provided by the invention, port includes the front end for connecting presynaptic neuron
Mouth Ipre, connection postsynaptic neuron rear port Ipost, synapse weight acquisition reading port IreadAnd N number of control port Ic,
Wherein, N is natural number.Front port IprePresynaptic neuron is connected, the cynapse is transferred to for receiving presynaptic neuron
Pulse signal, rear port IpostPostsynaptic neuron is connected, the pulse of the cynapse is transferred to for receiving postsynaptic neuron
Signal, control terminal IcMake front port I after receiving signalpreInput signal changes to electric conductivity value control law or rear end
Mouth IpostInput signal changes to electric conductivity value control law, reads port IreadIts electricity after receiving the signal less than threshold value
It leads value to remain unchanged, reads port IreadFor according to front port IpreInput signal, rear port IpostInput signal and control
End I processedcInput signal output signal output, output signal is related with the electric conductivity value of multiterminal artificial synapse device.Read port
IreadLevel V is read receivingreadAfterwards, reading device electric conductivity value, the weighted value as cynapse device.
Above-mentioned multiterminal artificial synapse device should have following electrology characteristic:Device electric conductivity value is non-volatile, i.e., electric conductivity value is each
It will not be lost after the power-off of port, after being once again powered up, device electric conductivity value is the electric conductivity value of power-cut time, and the electric conductivity value has in outside
It can change under effect excitation;Device electric conductivity value can pass through tune in the variation range of highest electric conductivity value and minimum electric conductivity value
The size of control external drive is continuously regulated and controled, and the synaptic weight value of multiterminal artificial synapse reads end by the synapse weight acquisition of device
Electric conductivity value indicate.
Based on this, the principle for defining the signal application port of above-mentioned multiterminal artificial synapse device is:If device is in outer end
Electric conductivity value becomes larger under mouth effectively positive incentive, then is the front port I for connecting presynaptic neuron by the port definitionpre;If device
Electric conductivity value becomes smaller under effective positive incentive of outside port, then is the rear port for connecting postsynaptic neuron by the port definition
Ipost;If the electric conductivity value of device does not change when outside port applies the voltage less than some threshold voltage, threshold value electricity
It is determined at all according to the characteristic of device, then the port can be defined as reading port Iread, read port IreadFor exporting output letter
Number, current value of the device under the port is collected with voltage value, then can obtain the electric conductivity value of device;If device is in outer end
Under the voltage effect of mouth, electric conductivity value is by front port Ipre, rear port IpostThe degree or rule of regulation and control can change, alternatively,
The outside port can assist the reading of electric conductivity value, then the port can be defined as the control of pilot controller part electric conductivity value adjusting
Port Ic。
Multiterminal artificial synapse device provided by the invention is equipped with front port Ipre, rear port Ipost, read port IreadAnd control
Port I processedc.Wherein, front port Ipre, rear port IpostIt is respectively used to receive the signal and postsynaptic neuron of presynaptic neuron
Signal, under the excitation of above-mentioned input signal, the electric conductivity value of device changes, and via read port IreadOr it reads
Port IreadWith control port IcElectric conductivity value is read after cooperation, realizes that the multiterminal artificial synapse device receives signal and output signal
Parallel.
For above-mentioned multiterminal artificial synapse device, a kind of modulation system of present invention offer is realized to multiterminal artificial synapse device
Carry out write operation:
Initialization operation is carried out to device, device is made to be in controllable state.Multiterminal artificial synapse device after initialization
The front port I of partpreInput direct-current level signal Vwrite, by the rear port I of the multiterminal artificial synapse device after initializationpostIt connects
Ground is realized and carries out write operation to artificial synapse device, and after a write operation, the electric conductivity value of device is increased to from controllable state
High conductivity state GH.Port I is read in devicereadApply and reads signal Vread, and read signal VreadThe conductance of device is not changed
Value, and acquire and read port IreadElectric current calculates the electric conductivity value of device.
By above-mentioned modulation system, the artificial synapse device can be made to be in high conductivity state GH, electric current handling capacity is big, can
The situation that biological synapse bonding strength is very high, synapse weight is big is simulated using the device under in this state.
For above-mentioned multiterminal artificial synapse device, the present invention provides another modulation system and realizes to multiterminal artificial synapse device
Part carries out write operation, and with above-mentioned write operation, there are following differences:The front end of multiterminal artificial synapse device after initialization
Mouth IpreInput write-in threshold value pulse Pwrite, write-in threshold value pulse PwriteAmplitude and width can make the conductance of cynapse device
Value increases supreme conductivity state G from controllable stateH。
For above-mentioned multiterminal artificial synapse device, a kind of modulation system of present invention offer is realized to multiterminal artificial synapse device
Carry out erasing operation:
Initialization operation is carried out to device, device is made to be in controllable state.Multiterminal artificial synapse device after initialization
The rear port I of partpostInput direct-current level signal Verase, by the front port I of the multiterminal artificial synapse device after initializationpreIt connects
Ground is realized and carries out erasing operation to artificial synapse device.After the erase operation, the electric conductivity value of device is decreased to from controllable state
Low conductivity state GL.Port I is read in devicereadApply and reads signal Vread, and read signal VreadThe conductance of device is not changed
Value, and acquire and read port IreadElectric current calculates the electric conductivity value of device.
By aforesaid operations, the artificial synapse device can be made to be in low conductivity state GL, electric current handling capacity is small, can be used
Device under in this state simulates the situation that biological synapse bonding strength is very low, synapse weight is small.
For above-mentioned multiterminal artificial synapse device, the present invention provides another modulation system and realizes to multiterminal artificial synapse device
Part carries out erasing operation, and with above-mentioned write operation, there are following differences:The rear end of multiterminal artificial synapse device after initialization
Mouth IpostInput erasing threshold value pulse Perase, erasing threshold value pulse PeraseAmplitude and width can make the conductance of cynapse device
Value is decreased to low conductivity state G from controllable stateL。
Based on the above-mentioned write operation to multiterminal artificial synapse device and erasing operation, under type such as may be used and realize device
The multistage adjusting of part electric conductivity value realizes under external electrical signals excitation, to the synaptic weight value of multiterminal artificial synapse device into
The multistage adjusting of row:
In front port IpreInput pulse signal sequence, rear port IpostGround connection ensures cynapse device in presynaptic nerve
Under the independent stimulation of first input pulse signal, the weighted value of the cynapse device will increase, each arteries and veins in control wave sequence
Signal is rushed, each pulse signal makes the electric conductivity value of artificial synapse device increase.It constantly repeats to apply pulse, applies pulse each time
Parameter can remain unchanged or suitably increase amplitude or pulsewidth, ensure that the application of each subpulse can make the electricity of device
Value increase is led, realizes that the increased degree of electric conductivity value of artificial synapse device becomes larger.The conductance of reading device after applying per subpulse
Value, since the electric conductivity value of device is non-volatile, the accumulation application of each subpulse can all make the electric conductivity value of device compared to the secondary arteries and veins
Punching increased before applying, and until reaching maximum conductance value, realize the regulation and control that device electric conductivity value continuously enlarges, i.e. the cynapse device
Weight can continuously increase multistagely, to realize enhanced synaptic plasticity function.
In rear port IpostInput pulse signal sequence, front port IpreGround connection ensures to input arteries and veins in postsynaptic neuron
It rushes under the independent stimulation of signal, the weighted value of the cynapse device can reduce, each pulse signal in control wave sequence, often
A pulse signal makes the electric conductivity value of artificial synapse device reduce.It constantly repeats to apply pulse, the parameter for applying pulse each time can
To remain unchanged or suitably increase amplitude or pulsewidth, ensure that the application of each subpulse can be such that the electric conductivity value of device subtracts
It is small, realize that the degree that the electric conductivity value of artificial synapse device reduces becomes larger.The electric conductivity value of reading device after applying per subpulse, due to
The electric conductivity value of device is non-volatile, before the accumulation of each subpulse applies the electric conductivity value of device can all applied compared to the subpulse
Reduced, until reaching minimum electric conductivity value, realize the regulation and control that device electric conductivity value is continuously reduced, is i.e. the cynapse device weight can be with
Continuously reduce multistagely, to realize suppressive synaptic plasticity function.
By applying an amplitude or pulsewidth less than threshold value in the presynaptic of above-mentioned artificial device or postsynaptic input terminal
Pulse (write-in threshold value pulse PwriteOr erasing threshold value pulse Perase) pulse signal PM, the other end is grounded, will can manually be dashed forward
Tentaculum part regulates and controls to intermediate conductivity state GM;Apply conductance reading level by reading end in the figure of the artificial synapse device
Vread, and reading signal VreadThe electric conductivity value of device will not be made to change;Collection terminal electric current calculates the electric conductivity value of device.
By changing pulse signal PMPulse parameter amplitude and pulsewidth size, which can be regulated and controled to
Different intermediate conductance state GMX, realize the multistage regulation and control of conductivity state;When the device is in different intermediate conductance state GMXWhen,
Electric current is different by the ability of the cynapse device, intermediate conductance state GMXUnder artificial synapse device can be used for simulate biological synapse
Different bonding strength, you can to simulate different biological synapse weighted values.
By above-mentioned regulation and control flow, you can with by regulating and controlling presynaptic or postsynaptic input signal, by the artificial synapse device
Part is configured to different stabilization conductivity states, to realize enhanced and suppressive synaptic plasticity function.
Based on the above-mentioned multistage adjusting of electric conductivity value to multiterminal artificial synapse device, may be used as under type realizes device
Pulse sequence relies on synaptic plasticity function:
(1) the front port I of presynaptic neuron is connected in the artificial synapse devicepreInput presynaptic pulse signal Ppre,
The rear port I of postsynaptic neuron is connected in the artificial synapse devicepostInput postsynaptic pulse signal Ppost;Control the presynaptic
The time difference Δ t of pulse signal and postsynaptic pulse signal is more than zero, and adjusts presynaptic pulse signal and believe with cynapse afterpulse
Number shape, obtain superposed signal after presynaptic pulse signal is superimposed with postsynaptic pulse signal so that superposed signal is just
It is more than negative sense maximum voltage value to maximum voltage value, effective amplitude of superposed signal is just and effective amplitude of the superposed signal
Size increases with Δ t and is reduced.Wherein, effective amplitude refers to the difference of superposed signal maximum generating positive and negative voltage value.When presynaptic arteries and veins
When rushing signal prior to the arrival cynapse of postsynaptic pulse signal, the signal that cynapse receives is equal to from front port IpreApply positive
The electric conductivity value of excitation, device increases, therefore synapse weight increases;By applying conductance in the reading port of the artificial synapse device
Read level Vread, collection terminal electric current calculates the electric conductivity value of device;Reading signal VreadThe electric conductivity value of device will not be made to send out
Changing.
When the time difference Δ t of presynaptic pulse signal and postsynaptic pulse signal increases, the superposed signal it is effective
Amplitude size increases with Δ t and is reduced, and so that the degree that the synapse weight of the artificial synapse device enhances is reduced, i.e., weight increases width
Degree is related with Δ t, and increasing degree reduces as Δ t increases, and meets relationship as shown in Figure 2, the i.e. enhancing of device synapse weight
The time difference of degree and presynaptic afterpulse signal meets the relationship of exponential decrease, simulates enhanced prominent in biological neural cynapse
Touch plastic sexual function.
(2) the front port I of postsynaptic neuron is connected in the artificial synapse devicepreInput presynaptic pulse signal Ppre,
The rear port I of postsynaptic neuron is connected in the artificial synapse devicepostInput postsynaptic pulse signal Ppost;Control the presynaptic
The time difference Δ t of pulse signal and postsynaptic pulse signal is less than zero, and adjusts presynaptic pulse signal and believe with cynapse afterpulse
Number shape, the superposed signal after making presynaptic pulse signal be superimposed with postsynaptic pulse signal, positive maximum voltage value is small
In negative sense maximum voltage value, effective amplitude of superposed signal be it is negative, and effective amplitude size of the superposed signal increase with Δ t and
Reduce, therefore, when postsynaptic pulse signal reaches cynapse prior to presynaptic pulse signal, the signal that cynapse receives is equal to
From rear port IpostApply positive incentive, the electric conductivity value of device reduces, and therefore synapse weight reduces, by the artificial synapse device
The reading end of part applies conductance and reads level Vread, collection terminal electric current calculates the electric conductivity value of device;Reading signal VreadNo
The electric conductivity value of device can be made to change.
When the time difference Δ t increases of presynaptic pulse signal and postsynaptic pulse signal, and the superposed signal is effective
Amplitude size increases with Δ t and is reduced, and so that the degree that the synapse weight of the artificial synapse device reduces is reduced, i.e., weight reduces width
Degree is related with Δ t, reduces amplitude and reduces as Δ t increases, meets relationship as shown in Figure 2, the i.e. reduction of device synapse weight
The time difference of degree and presynaptic afterpulse signal meets the relationship of exponential decrease;It is prominent to simulate suppressive in biological neural cynapse
Touch plastic sexual function.
Based on above-mentioned artificial synapse device conductance can it is multistage regulation and control to realize enhancing and suppressive synaptic plasticity function
Characteristic, design device connection presynaptic neuron waveform input signal and device connection postsynaptic neuron input signal
Waveform controls artificial synapse device synapse weight enhancing and inhibits to reach the time difference of cynapse with presynaptic afterpulse and change
Degree, in the artificial synapse device realize pulse sequence rely on synaptic plasticity.
The device that the artificial synapse device of above-mentioned multiterminal regulation and control is inputted as a kind of multiterminal, is provided simultaneously with two end artificial synapses
The electrology characteristic of device.Wherein, the artificial synapse device of above-mentioned multiterminal input can be two dimensional surface multiterminal entering apparatus and three
The three-dimensional multiterminal entering apparatus of dimension.3 D stereo multiterminal entering apparatus can be bipolar transistor devices, field-effect transistor device
Part and other three-dimensional structure multi-electrode entering apparatus.Above-mentioned 3 D stereo multiterminal input crystal tube device can be NOR Flash
Device, NAND Flash devices, SONOS devices, floating gate transistor device, ferro-electric field effect transistor device, spin transistor
Device or oxide crystal tube device etc..
The present invention is using four end floating gate transistor devices as embodiment one, to be specifically described the implementation of the present invention.Such as Fig. 3
It is shown, to realize that pulse sequence relies on the physical circuit figure of synaptic plasticity function based on floating transistor artificial synapse device.
Floating transistor includes the gate oxide 102 for stacking gradually arrangement, floating gate layer 103, floating boom insulating layer 104 and substrate layer, wherein
Floating boom insulating layer 104 is between floating gate layer 103 and substrate layer;Gate oxide 102 is equipped with grid 101 above, in substrate layer two
Side is equipped with drain electrode 105 and source electrode 106 successively, and body electrode 107 is equipped in substrate layer.The grid of floating transistor cynapse device
101, as the rear port I inputted for postsynaptic pulse signalpost, the body electrode 107 of floating transistor, as cynapse
The front port I of prepulse signal inputpre, the drain electrode 105 of floating transistor, the reading as multiterminal artificial synapse device conductance
Outlet Iread, as the weighted value of artificial synapse, 106 end of source electrode of floating transistor is as control terminal Ic, reading multiterminal people
It is grounded when work cynapse device electric conductivity value.
The knots modification of floating transistor electric conductivity value is the offset based on switching threshold voltage, and the offset of switching threshold
It is the quantity that electronics is stored based on floating gate layer, the more of floating gate layer storage electronics can be controlled by changing the size of external drive signal
Few, when to make the quantity for storing electronics in floating gate layer that multistage variation be presented, therefore multistage change is also presented in the electric conductivity value of floating gate layer
Change.
For the above-mentioned multiterminal artificial synapse device based on floating transistor, the present invention provides a kind of modulation system realization pair
Multiterminal artificial synapse device realizes read-write operation:
Apply forward voltage on 101 end of grid of floating transistor, initialization operation is carried out to floating transistor, makes to float
Electronics is accumulated and stored in the floating gate layer 103 of gate transistor, is in relatively low conductivity state.In floating transistor body electrode 107
Input direct-current level signal Vwrite, grid 101 is grounded, write operation is carried out to floating transistor, in DC level signal
VwriteUnder the action of, the charge that floating gate layer 103 stores is sucked out under the action of electric field from floating gate layer 103, the switching threshold of device
Threshold voltage reduces, and the electric conductivity value of device increases supreme conductivity state G from relatively low conductivity stateH。
The acquisition mode of device electric conductivity value is:Apply what additional device electric conductivity value was read in the grid 101 of floating transistor
Bias voltage Vb, apply in the drain electrode 105 of floating transistor and read signal Vread, collection terminal electric current calculates the conductance of device
Value.The artificial synapse device is in high conductivity state GHWhen, electric current handling capacity is big, the device under can be used in this state
To simulate the situation that biological synapse bonding strength is very high, synapse weight is big.
It is above-mentioned in 107 input direct-current level signal V of floating transistor body electrodewriteIt could alternatively be write-in threshold value pulse
Pwrite。
For the above-mentioned multiterminal artificial synapse device based on floating transistor, the present invention provides a kind of modulation system realization pair
Multiterminal artificial synapse device realizes erasing operation:
Apply forward voltage in the body electrode 107 of floating transistor, initialization operation is carried out to floating transistor, makes floating boom
The electronics accumulated in transistor is sucked out completely, is in relatively high conductivity state.In 101 input direct-current of grid of floating transistor
Level signal Verase, the body electrode 107 of floating transistor is grounded, erasing operation is carried out to artificial synapse device, in direct current
Ordinary mail VeraseUnder the action of, electronics is injected in floating gate layer, after level signal is removed, electrons are stored in floating gate layer, device
Switching threshold voltage will increase, the electric conductivity value of device is reduced from relatively high conductivity state to low conductivity state GL。
The acquisition mode of device electric conductivity value is identical with read-write operation, which is in low conductivity state GL
When, electric current handling capacity is small, and biological synapse bonding strength is very low, cynapse power to simulate for the device under can be used in this state
The small situation of weight.
It is above-mentioned in 107 input direct-current level signal V of floating transistor body electrodeeraseIt could alternatively be erasing threshold value pulse
Perase。
For the above-mentioned multiterminal artificial synapse device based on floating transistor, the present invention provides a kind of modulation system realization pair
Multiterminal artificial synapse device, which is realized, carries out multistage conductance regulation and control behaviour:
By in the floating transistor body electrode 107 of above-mentioned high conductance or low conductivity state or the grid of floating transistor
101 apply an amplitude or pulsewidth less than threshold value pulse (PwriteOr Perase) pulse signal PM, the other end is grounded, it can be by people
Work cynapse device regulates and controls to intermediate conductivity state GM;By changing pulse signal PMPulse parameter amplitude, pulsewidth and number of pulses
Size, which can be regulated and controled to different intermediate conductance state GMX, wherein X is natural number, realizes conductivity state
Multistage regulation and control.
Specifically, for increasing pulse amplitude, realize that the increased mode of multistage conductance is in the device:
As shown in figure 4, in the constantly increased pulse train of 107 input pulse amplitude of body electrode of floating transistor, will float
The grid 101 of gate transistor is grounded, and front and back pulse signal is superposed to positive incentive, that is, is equal to the body electricity in floating transistor
Pole 107 inputs positive effectively excitation, often applies a pulse, is all once read to the electric conductivity value of device, as shown in fig. 6,
In this operation, as each subpulse applies, the electric conductivity value of device can increased before applying compared to the subpulse, electric
It leads value to apply with each subpulse and increase, synapse weight enhancing.
Equally for increasing pulse amplitude, realize that the mode that multistage conductance reduces is in the device:
As shown in figure 5, the body electrode 107 of floating transistor is grounded, in the 101 input pulse width of grid of floating transistor
Value constantly increased pulse train, front and back pulse signal are superposed to negative incentive, that is, are equal to the grid in floating transistor
The positive effectively excitation of 101 inputs, often applies a pulse, is all once read to the electric conductivity value of device, such as Fig. 6, in the behaviour
In work, as each subpulse applies, the electric conductivity value of device can be reduced before applying compared to the subpulse, electric conductivity value with
Each subpulse applies and reduces, and synapse weight weakens.
In the mistake for carrying out multistage conductance regulation and control to the multiterminal artificial synapse device based on floating transistor according to the method described above
Cheng Zhong can change 107 input signal of body electrode or grid 101 be defeated in the 106 end input signal of source electrode of floating transistor
Enter electric conductivity value control law of the signal to floating transistor.
Fig. 6 then illustrates the multistage regulation and control result figure of conductance of device.When the device is in different intermediate conductance state GMXWhen,
Electric current is different by the ability of the cynapse device, intermediate conductance state GMXUnder artificial synapse device can be used for simulate biological synapse
Different bonding strength, you can to simulate different biological synapse weighted values;When device is adjusted supreme electricity by low conductivity state
When leading state, shows that the synapse weight of the artificial synapse device increases, can be used to simulate enhanced synaptic plasticity behavior;And work as
When device is adjusted by high conductivity state to low conductivity state, shows that the synapse weight of the artificial synapse device reduces, can be used to
Simulate suppressive synaptic plasticity behavior.
Based on the embodiment of floating transistor artificial synapse device, a kind of artificial synapse inputted based on the multiterminal is provided
The pulse sequence of device relies on synaptic plasticity function and realizes and regulate and control method, specifically comprises the following steps:
As shown in fig. 7, the shape of the 101 equal input pulse of grid in the body electrode 107 and floating transistor of floating transistor
Shape is triangular pulse signal, and the grid 101 of the 107 input pulse signal of body electrode and floating transistor that make floating transistor is defeated
The time difference for entering pulse signal is more than zero, makes the grid of body electrode 107 the input pulse signal and floating transistor of floating transistor
Effective amplitude is just waveform forward direction partial amplitude A after being superimposed after 101 input pulse Signal averaging of pole1More than negative-going portion width
Value A2, the electric conductivity value increase of device, simulates enhanced synaptic plasticity function in biological neural cynapse at this time.
The grid 101 of triangular pulse signal and floating transistor is inputted by adjusting the body electrode 107 of floating transistor
Triangular pulse signal is inputted, the grid 101 of the 107 input pulse signal of body electrode and floating transistor that make floating transistor is defeated
The time difference for entering pulse signal is less than zero, and effective amplitude is negative after superposition, that is, waveform forward direction partial amplitude A after being superimposed1Less than negative
To partial amplitude A2, the electric conductivity value reduction of device, simulates suppressive synaptic plasticity function in biological neural cynapse at this time.
As shown in Figure 9 and as shown in Figure 10, floating transistor artificial synapse device is inputted based on four end, when realizing pulse
When sequence dependent form synaptic plasticity, triangular pulse signal can also be replaced with to square-wave signal, the body electrode of floating transistor
The grid 101 of 107 input square-wave pulse signals and floating transistor inputs time difference, waveform after superposition between square-wave pulse signal
Magnitude relation with triangular pulse signal as input signal when it is identical.I.e. as time difference Δ t > 0, waveform is positive after superposition
Partial amplitude A1More than negative-going portion amplitude A2;As time difference Δ t < 0, waveform forward direction partial amplitude A after superposition1Less than negative sense
Partial amplitude A2。
The present invention is using four end iron field effect transistor devices as embodiment two, is that four ends input as shown in figure 11
Ferro-electric field effect transistor device.Ferro-electric field effect transistor includes ferroelectric material layer 202, insulating layer 203 and substrate layer, absolutely
Edge layer 203 is provided with grid 201, in substrate between ferroelectric material layer 202 and substrate layer above ferroelectric material layer 202
Layer is provided at both ends with drain electrode 204 and source electrode 205, and body electrode 206 connects substrate layer.
The operation principle of ferro-electric field effect transistor is:When adding positive voltage at 201 end of grid, 206 end of body electrode is grounded
When, write operation can be carried out to device, can induce electronics due to iron electric polarization effect, in the conducting channel of source and drain interpolar makes
Channel current rises, and the electric conductivity value of device increases;When applying positive voltage at 206 end of body electrode, and 201 end of grid being grounded,
Erasing operation can be carried out to device, electronics is discharged in raceway groove, accumulates positive charge, and source and drain electrode current can be truncated, source-drain electrode
Electric current declines, and the electric conductivity value of device reduces.
Based on the operation principle, 201 end of grid of ferro-electric field effect transistor can be defined as to synapse transistor company
Meet the I of presynaptic neuronpre206 end of body electrode is defined as the I that the synapse transistor connects postsynaptic neuron by endpost
End.The drain electrode 204 of ferro-electric field effect transistor, the conductance as the cynapse device based on ferro-electric field effect transistor read end
Iread, the weighted value of artificial synapse device, the source electrode 205 of ferro-electric field effect transistor, as control terminal I can be readc, iron
The source electrode 205 of field effect transistor is grounded in reading device electric conductivity value.
The control manipulation principle of multistage conductance based on embodiment two is the same as embodiment one.
Pulse sequence based on embodiment two relies on the control manipulation principle of synaptic plasticity with embodiment one.
The present invention is using SONOS devices as embodiment three, as shown in figure 12, for the SONOS devices of four ends input.
SONOS devices include the upper oxidation insulating layer 302 for stacking gradually placement, silicon nitride layer 303, lower oxidation insulating layer 304 and substrate
Layer, lower oxidation insulating layer 304 are provided at both ends with drain electrode 305 and source between silicon nitride layer 303 and substrate layer, in substrate layer
Electrode 306, body electrode 307 connect substrate layer.
The operation principle of SONOS devices is:Apply positive voltage on 301 end of grid, it, can be with when body electrode 307 is grounded
Erasing operation is carried out to device, during being somebody's turn to do, by injection electronics in 302 end of silicon nitride layer, the switching threshold voltage of device is corresponding
Increase, drifts about to the right, under same grid voltage, raceway groove current value becomes smaller between source-drain electrode, and device electric conductivity value accordingly becomes smaller;In body electrode
307 ends apply positive voltage, and 301 end of grid is grounded, and can carry out write operation to device, during being somebody's turn to do, 302 end of silicon nitride layer
The electronics of middle accumulation is sucked out, and the switching threshold of device accordingly reduces, and drifts about to the left, under same gate voltage, source and drain interpolar
Channel current value becomes larger, and device electric conductivity value accordingly increases.
Based on the operation principle, the body electrode 307 of SONOS devices can be defined as to the synapse transistor connection presynaptic
The I of neuronpreGrid 301 is defined as the I that the synapse transistor connects postsynaptic neuron by endpostEnd, SONOS devices
Drain electrode 305, the conductance as SONOS cynapse devices read end Iread, the weighted value of the artificial synapse device can be read,
The source electrode 306 of SONOS devices is grounded in reading device electric conductivity value, as control terminal Ic。
The control manipulation principle of multistage conductance based on embodiment three is the same as embodiment one.
Pulse sequence based on embodiment three relies on the control manipulation principle of synaptic plasticity with embodiment one.
In the present invention, described multi-terminal device, floating transistor artificial synapse device including but not limited in embodiment,
Ferro-electric field effect transistor artificial synapse device and SONOS artificial synapse devices, floating transistor artificial synapse device, ferroelectricity field
Effect transistor artificial synapse device and SONOS artificial synapse devices are only presently preferred embodiments of the present invention;Similarly, it is based on
In the embodiment of floating gate transistor device, the waveform of the connection type and signal of external circuit and device interface is this hair
Bright preferred embodiment, any all any modification, equivalent and improvement done within spirit of that invention and principle,
It should be included within the scope of the present invention.
Claims (9)
1. a kind of multiterminal artificial synapse device, which is characterized in that the multiterminal artificial synapse device is electric conductivity value non-volatile device,
Device electric conductivity value changes in the case where external drive continuously regulates and controls, and device includes:
The front port I for making its electric conductivity value become larger after receiving signalpre, for connecting presynaptic neuron, receive presynaptic nerve
Member transmission signal;
The rear port I for making its electric conductivity value become smaller after receiving signalpost, for connecting postsynaptic neuron, receive postsynaptic neuronal
Member transmission signal;
Make front port I after receiving signalpreInput signal changes to electric conductivity value control law or rear port IpostInput
Signal control terminal I changed to electric conductivity value control lawc, for receiving control signal;And
The reading port I that its electric conductivity value remains unchanged after receiving the signal less than threshold valueread, for according to front port IpreIt is defeated
Enter signal, rear port IpostInput signal and control terminal IcInput signal output signal output, output signal is artificial with multiterminal
The electric conductivity value of cynapse device is related.
2. multiterminal artificial synapse device as described in claim 1, which is characterized in that the multiterminal artificial synapse device includes
Multiple control terminal Ic。
3. multiterminal artificial synapse device as claimed in claim 1 or 2, which is characterized in that the multiterminal artificial synapse device is
Two dimensional surface device or three-dimensional structure device.
4. multiterminal artificial synapse device as claimed in claim 3, which is characterized in that three-dimensional structure device is bipolar transistor
Device or FET device.
5. multiterminal artificial synapse device as claimed in claim 4, which is characterized in that FET device NOR
Flash devices, NAND Flash devices, SONOS devices, floating gate transistor device, ferro-electric field effect transistor device, spin traps
Field effect transistor devices or oxide field-effect transistor device.
6. such as multiterminal artificial synapse device described in any one of claim 1 to 5, which is characterized in that when multiterminal artificial synapse device
When part is floating gate transistor device or SONOS devices, rear port I of the grid as multiterminal artificial synapse devicepost, body electrode
Front port I as multiterminal artificial synapse devicepre, reading port I of the drain electrode as multiterminal artificial synapse deviceread, source electricity
Control terminal I extremely as multiterminal artificial synapse devicec;
Wherein, floating transistor includes the gate oxide for stacking gradually arrangement, floating gate layer, floating boom insulating layer and substrate layer;Grid oxygen
Change and be equipped with grid above layer, drain electrode and source electrode are equipped with successively in substrate layer both sides, body electrode is equipped in substrate layer;
SONOS devices include the upper oxidation insulating layer for stacking gradually placement, silicon nitride layer, lower oxidation insulating layer and substrate layer;On
It is equipped with grid above oxidation insulating layer, is provided at both ends with drain electrode and source electrode in substrate layer, body electrode connects substrate layer.
7. such as multiterminal artificial synapse device described in any one of claim 1 to 5, which is characterized in that when multiterminal artificial synapse device
When part is ferro-electric field effect transistor, front port I of the grid as multiterminal artificial synapse devicepre, body electrode is artificial as multiterminal
The rear port I of cynapse devicepost, reading port I of the drain electrode as multiterminal artificial synapse deviceread, source electrode end is as multiterminal
The control terminal I of artificial synapse devicec;
Ferro-electric field effect transistor includes the ferroelectric material layer, insulating layer and substrate layer for stacking gradually arrangement, in ferroelectric material layer
Top is provided with grid, is provided at both ends with drain electrode and source electrode in substrate layer, body electrode connects substrate layer.
8. the multistage synaptic plasticity modulator approach of multiterminal artificial synapse device as described in any one of claim 1 to 7, special
Sign is, includes the following steps:
In front port IpreInput pulse signal sequence, by rear port IpostIt is grounded, each pulse signal energy in sequences of pulsed signals
The electric conductivity value of artificial synapse device is set to increase, and the amplitude of the latter pulse signal is more than previous pulse in sequences of pulsed signals
The amplitude of signal or the width of the latter pulse signal are more than the width of previous pulse signal, make artificial synapse device
The increased degree of electric conductivity value becomes larger, and realizes enhanced synaptic plasticity function;Or
In rear port IpostInput pulse signal sequence, by front port IpreIt is grounded, each pulse signal energy in sequences of pulsed signals
The electric conductivity value of artificial synapse device is set to reduce, and the amplitude of the latter pulse signal is more than previous pulse in sequences of pulsed signals
The amplitude of signal or the width of the latter pulse signal are more than the width of previous pulse signal, make artificial synapse device
The degree that electric conductivity value reduces becomes larger, and realizes suppressive synaptic plasticity function.
9. the pulse sequence of multiterminal artificial synapse device as described in any one of claim 1 to 7 relies on synaptic plasticity modulation
Method, which is characterized in that include the following steps:
Front port IpreInput presynaptic pulse signal Ppre, rear port IpostInput postsynaptic pulse signal Ppost;Cynapse prepulse
The time difference Δ t of signal and postsynaptic pulse signal is more than zero, and effective amplitude of superposed signal is more than zero, and superposed signal has
Poor Δ t increases and reduces effect amplitude at any time;Realize that enhanced pulse sequence relies on synaptic plasticity function;Or
Front port IpreInput presynaptic pulse signal Ppre, rear port IpostInput postsynaptic pulse signal Ppost;Cynapse prepulse
The time difference Δ t of signal and postsynaptic pulse signal is less than zero, and effective amplitude of superposed signal is less than zero, and superposed signal has
Poor Δ t increases and reduces effect amplitude at any time, realizes that suppressive pulse sequence relies on synaptic plasticity function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810516810.0A CN108777153B (en) | 2018-05-25 | 2018-05-25 | Multi-terminal input synapse device and plasticity modulation method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810516810.0A CN108777153B (en) | 2018-05-25 | 2018-05-25 | Multi-terminal input synapse device and plasticity modulation method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108777153A true CN108777153A (en) | 2018-11-09 |
CN108777153B CN108777153B (en) | 2021-01-26 |
Family
ID=64027597
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810516810.0A Active CN108777153B (en) | 2018-05-25 | 2018-05-25 | Multi-terminal input synapse device and plasticity modulation method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108777153B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112381217A (en) * | 2020-11-06 | 2021-02-19 | 北京大学 | Method for realizing excitation and inhibition functions of synapse |
CN113646912A (en) * | 2020-02-27 | 2021-11-12 | Tdk株式会社 | Arithmetic circuit and neuromorphic device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103078055A (en) * | 2013-01-04 | 2013-05-01 | 华中科技大学 | Unit, device and method for simulating biological neuronal synapsis |
US20170068885A1 (en) * | 2013-08-02 | 2017-03-09 | International Business Machines Corporation | Dual deterministic and stochastic neurosynaptic core circuit |
JP2017097693A (en) * | 2015-11-26 | 2017-06-01 | Kddi株式会社 | Data prediction device, information terminal, program, and method performing learning with data of different periodic layer |
CN107454965A (en) * | 2015-05-21 | 2017-12-08 | 谷歌公司 | Batch processing in neural network processor |
-
2018
- 2018-05-25 CN CN201810516810.0A patent/CN108777153B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103078055A (en) * | 2013-01-04 | 2013-05-01 | 华中科技大学 | Unit, device and method for simulating biological neuronal synapsis |
US20170068885A1 (en) * | 2013-08-02 | 2017-03-09 | International Business Machines Corporation | Dual deterministic and stochastic neurosynaptic core circuit |
CN107454965A (en) * | 2015-05-21 | 2017-12-08 | 谷歌公司 | Batch processing in neural network processor |
JP2017097693A (en) * | 2015-11-26 | 2017-06-01 | Kddi株式会社 | Data prediction device, information terminal, program, and method performing learning with data of different periodic layer |
Non-Patent Citations (1)
Title |
---|
ROSHAN GOPALAKRISHNAN等: "Triplet Spike Time-Dependent Plasticity in a Floating一Gate Synapse", 《IEEE TRANSACTIONS ON NEURAL NETWOKRS AND LEARNING SYSTEMS》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113646912A (en) * | 2020-02-27 | 2021-11-12 | Tdk株式会社 | Arithmetic circuit and neuromorphic device |
CN113646912B (en) * | 2020-02-27 | 2024-02-27 | Tdk株式会社 | Arithmetic circuit and neuromorphic device |
CN112381217A (en) * | 2020-11-06 | 2021-02-19 | 北京大学 | Method for realizing excitation and inhibition functions of synapse |
CN112381217B (en) * | 2020-11-06 | 2022-05-20 | 北京大学 | Method for realizing excitation and inhibition functions of synapse |
Also Published As
Publication number | Publication date |
---|---|
CN108777153B (en) | 2021-01-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107068708B (en) | A kind of floating gate memristor | |
Kim et al. | Silicon synaptic transistor for hardware-based spiking neural network and neuromorphic system | |
He et al. | Recent progress on emerging transistor‐based neuromorphic devices | |
CN109447250B (en) | Artificial neuron based on battery effect in memristor | |
Likharev | CrossNets: Neuromorphic hybrid CMOS/nanoelectronic networks | |
Hu et al. | Memristive crossbar array with applications in image processing | |
CN110289317B (en) | Ferroelectric graphene transistor, complementary type synapse device based on ferroelectric graphene transistor and regulation and control method | |
Mostafa et al. | Implementation of a spike-based perceptron learning rule using TiO2− x memristors | |
WO2016069334A1 (en) | Neuromorphic synapses | |
CN108777153A (en) | A kind of its plasticity modulator approach of multiterminal input cynapse device | |
Subramaniam et al. | Spike-timing-dependent plasticity using biologically realistic action potentials and low-temperature materials | |
Ziegler et al. | Mimic synaptic behavior with a single floating gate transistor: A MemFlash synapse | |
CN103346256A (en) | Memristor based on ferroelectric tunnel junction | |
CN109787592B (en) | Random nerve pulse generator | |
CN106847817A (en) | Artificial synapse device and its preparation method and application with multiterminal configuration | |
Gerasimova et al. | Design of memristive interface between electronic neurons | |
Sun et al. | Advanced synaptic devices and their applications in biomimetic sensory neural system | |
Xue et al. | Native drift and Mott nanochannel in layered V2O5 film for synaptic and nociceptive simulation | |
Zhao et al. | Recent progress on neuromorphic synapse electronics: from emerging materials, devices, to neural networks | |
Hsu | Dendritic computation and plasticity in neuromorphic circuits | |
Li et al. | Flexible and Compatible Synaptic Transistor Based on Electrospun In 2 O 3 Nanofibers | |
Chen et al. | Perspective on oxide-based three-terminal artificial synapses in physical neural networks | |
CN109935686A (en) | Memory resistor and preparation method thereof with double pulses laser characteristic | |
Cederström et al. | A model based comparison of BiFeO3 device applicability in neuromorphic hardware | |
CN109346599A (en) | More iron tunnel knot memristors and preparation method thereof |
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 |