CN107085628A - A kind of adjustable weights modular simulation method of cell neural network - Google Patents
A kind of adjustable weights modular simulation method of cell neural network Download PDFInfo
- Publication number
- CN107085628A CN107085628A CN201710169625.4A CN201710169625A CN107085628A CN 107085628 A CN107085628 A CN 107085628A CN 201710169625 A CN201710169625 A CN 201710169625A CN 107085628 A CN107085628 A CN 107085628A
- Authority
- CN
- China
- Prior art keywords
- memristor
- simulink
- cell
- neural network
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
-
- 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
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Biophysics (AREA)
- Neurology (AREA)
- Geometry (AREA)
- Molecular Biology (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of adjustable weights modular simulation method of cell neural network, it is mainly based upon memristor and builds memristor cynapse bridge circuit and is emulated, by adjusting size and the time of memristor bridge circuit input pulse, the cell neural network Weight template of needs can be accurately obtained.The present invention can realize the adjustable weights module of cell neural network.
Description
Technical field
The present invention relates to neutral net and the interdisciplinary field of memristor, more particularly to a kind of cell neural network are adjustable
Weights modular simulation method.
Background technology
Cell neural network (Cellular Neural Network, CNN) is that Cai Shaotang professors put forward in 1988
A kind of information processing system.It is much like with neutral net, cell neural network be it is a kind of can process signal in real time it is extensive
Non-linear analog circuit.The basic structure circuit unit of cell neural network is referred to as cell, including linear processes circuit
Original paper, generally linear electric capacity, linear resistance, linear processes controlled source, and independent source.The structure of cell neural network is very
As cellular automaton, that is, any cell are only connected with its flanking cell.Flanking cell can be directly affected, not straight
Connecing connected cell may be affected one another by the continuous dynamic communication of network.
The structure of cell neural network is as shown in figure 1, C (i, j) represents the cell of the i-th row jth row, in cell neural network
In, the structure of all cells is all identical.Possess in network (2r+1)2The cell in field is referred to as internal cell, remaining
It is then border cell.Cell neural network has Control architecture and feedback template, and Control architecture determines input signal to cell
The influence degree of state, feedback template then determines influence degree of the output to cell C (i, j) of flanking cell.Because cell
Neutral net is based on circuit realiration, so to specific neutral net, its template weights is fixed, is so limited
CNN versatility.For different applications, it is necessary to build the circuit structure of different Weight templates.
The cynapse of current cell neural network typically uses resistance, and electric capacity and transistor circuit are realized.It is real using resistance
Existing synapse cell is a kind of static behavior, i.e., it can not be changed again once this synaptic structure is prepared for, therefore only
It can apply to the scene of non-study.Capacity fall off characteristic cause it can only can of short duration preservation synaptic weight, if so using it
Realize that cynapse then needs dynamically to update synaptic weight according to certain frequency.Except electric capacity and resistance, at present using floating coral crystal
Pipe realizes that the trial of cynapse also achieves certain success with analog multiplier, but it is existed significantly in use
Non-linear behavior.
Memristor is a kind of variable Two-port netwerk circuit element of resistance, with resistance identical dimension.1971, Cai was few
Chinese bush cherry professor theoretically demonstrates the presence of memristor.2008, the researcher of HP Lab claimed in fact in Nature dispatches
A kind of memristor of titania structure is showed.With the discovery of Hewlett-Packard's memristor, cytocidal action is realized using memristor cynapse
The synaptic weight of network becomes a kind of new selection.Memristor is that a kind of resistance can be adjusted and storage information is non-volatile in itself
Element, so relative to resistance is used, the synaptic weight of the realization such as electric capacity realizes that the cell that can be adjusted is dashed forward using memristor
Tactile weights have very big application prospect.Memristor has and the very much like characteristic of synapse cell weights, you can with outer
Plus pulse and change and when it is additional stimulate disappear after can keep stateful constant.Realize that synapse cell is weighed using memristor
Value, it is possible to achieve weights can adjust, while the multiplier circuit of traditional synaptic weight circuit is no longer needed, so that net
Network scale down.But it is not found the adjustable weights modular simulation side of cell neural network based on memristor in the prior art
Method.
The content of the invention
Goal of the invention：There is provided a kind of adjustable weights mould of cell neural network for the problem of present invention exists for prior art
Block emulation mode.
Technical scheme：The adjustable weights modular simulation method of cell neural network of the present invention comprises the following steps：
Step 1：The Simulink models of memristor are built, are specifically included：
1-1：The integration to memristor state variable w (t) is realized using Simulink integration module, wherein solver is adopted
Use ode45；
1-2：Memristor window function is realized using Simulink SQL module
Wherein, D is the thickness parameter of memristor, and window function parameter p is 10；
1-3：Realized using Simulink Gain modulesWherein, uvFor drift speed, RONRepresent that memristor is complete
Resistance value when portion adulterates；
1-4：Calculate ROFF- w (t) * Δ R, are then sent to signal conversion module Simulink-Ps Convert by result
Be converted to and variable resistor module is input to after physical signalling, so as to realize the change of memristor resistance；Wherein, ROFFRepresent memristor
Resistance during device whole undoped, Δ R represents ROFF-RON；
Step 2：The signaling conversion circuit of cell neural network and memristor module is built, is specifically included：
Physical signalling is converted a signal into using Simulink-Ps convert, SimScape control is then inputted
Voltage source processed, you can to obtain SimScape voltage source pulses, be converted to Simulink voltage pulse signals so as to realize
The signal conversion of SimScape signals；
Step 3：In Simulink, by the way that memristor first is built into wheatstone bridge configuration by way of parallel connection is connected again
Memristor Sudden-touch circuit；
Step 4：It is integrated into after the memristor Sudden-touch circuit built using Simulink is encapsulated in cell neural network circuit,
Specially：Memristor bridge Sudden-touch circuit is packaged into the subsystem of a Two-port netwerk, an end using Simulink Subsystem
Mouth is the input of voltage source, and one is the current source output after synaptic weight.
Beneficial effect：Compared with prior art, its remarkable advantage is the present invention：(1) present invention is realized based on memristor
The adjustable weights module of cell neural network emulation；(2) compared to traditional Sudden-touch circuit once it is determined that cynapse can not be changed
The size of weights, memristor cynapse can be using being that synaptic weight size is accurately adjusted by voltage pulse；(3) memristor is one
Nanoscale circuit element is planted, synapse cell is realized using memristor, can greatly lift the integrated level of network；(4) use and recall
Resistance device realizes synapse cell, can remove the mlultiplying circuit in conventional cell Sudden-touch circuit, can reduce the scale of network；(5)
Memristor is a kind of non-volatile elements, i.e., its memristor value will not change in the case of no outside source, use
The synapse cell circuit that memristor is realized as the cynapse that electric capacity is realized without needing timing adjustment cynapse size；(6) memristor
Device is analogous to the element of resistance, and cell is realized using memristor, can greatly reduce the power consumption of circuit.
Brief description of the drawings
Fig. 1 is cell neural network structural representation；
Fig. 2 is that cell neural network and the signaling conversion circuit of memristor module are applied；
Fig. 3 is memristor Sudden-touch circuit figure；
Fig. 4 is cell neural network practical circuit diagram；
Fig. 5 is the weights variation diagram for the memristor for adjusting synapse cell to 1, and now the applying voltage pulse time is 0, because
Memristor resistance initial value is set to 1, so applying voltage pulse can essentially be not added with now；
Fig. 6 is the weights variation diagram for the memristor for adjusting synapse cell to 0.5, and the additional cycle is 5s, and dutycycle is %
10.57 voltage pulse, from the figure, it can be seen that memristor synaptic weight is precisely adjusted as required weights size；
Fig. 7 is the weights variation diagram for the memristor for adjusting synapse cell to 0, and the additional cycle is 5s, and dutycycle is %18.60
Voltage pulse.From the figure, it can be seen that memristor synaptic weight is precisely adjusted as required weights size；
Fig. 8 is the weights variation diagram for the memristor for adjusting synapse cell to 1, and the additional cycle is 5s, and dutycycle is %36.40
Voltage pulse, from the figure, it can be seen that memristor synaptic weight is precisely adjusted as required weights size；
Fig. 9 is the cell feedback template A figures built using memristor synaptic weight circuit, and wherein WeightA encapsulation is prominent
Touch weights circuit；
Figure 10 is the cell Control architecture B built using memristor synaptic weight circuit figure, and what wherein WeightB was encapsulated is
Synaptic weight circuit；
Figure 11 is adjusting template weights figure, and the Control architecture B that (a) is represented, what (b) was represented is feedback template A；
The initial value figure of each cell of network when Figure 12 (a) is emulation, (b) is the output figure of network；
Figure 13 is the last state diagram of each cell in network；
Figure 14 is that cell C (2,2) state is changed over time with output when being emulated using memristor cell neural network
Figure.
Embodiment
Cell neural network mathematical modeling
Cell C (i, j) Cellular Neural Networks are：
In formula, C is linear capacitance, node voltage vxijRepresent cell C (i, j) state, node voltage vuijRepresent cell C
The input of (i, j), node voltage vyijRepresent cell C (i, j) output, Nr(i, j) expression cell C (i, j) neighborhood, A (i,
j；K, l) be cell C (i, j) output feedback template, represent cell C (k, l) output and cell C (i, j) between connection weight
Value, B (i, j；K, l) be cell C (i, j) input Control architecture, represent cell C (k, l) input and cell C (i, j) between
Connection weight, R represents resistance, and I represents independent current.
Memristor mathematical modeling
Memristor is a kind of Two-port netwerk circuit element by Memorability, and it is defined as derivative of the magnetic flux to electric chargeM (t) is the resistance of memristor,Magnetic flux, q (t) is the quantity of electric charge, it can be seen that the resistance of memristor with
The integration of electric current has relation, so the Two-port netwerk element that memristor, which is a kind of resistance, to be adjusted, and its voltage-current relationship isThe resistance of memristor is
Wherein uVIt is drift speed, size is 10-10cm2s-1V-1.The state variable of memristor is typically nonlinear, that is, needs to be multiplied by
One window function Fp(w), it is used herein f (w)=1- (2w-1)2p, obtain
Emulation mode
Step 1：The Simulink models of memristor are built, are specifically included：
1-1：The integration to memristor state variable w (t) is realized using Simulink integration module, wherein solver is adopted
Use ode45；
1-2：Memristor window function is realized using Simulink SQL module
Wherein, D is the thickness parameter of memristor, and window function parameter p is 10；
1-3：Realized using Simulink Gain modulesIts size is 1e in this model4, wherein, uvFor drift
Move speed, RONRepresent resistance value when memristor all adulterates；
1-4：Calculate ROFF- w (t) * Δ R, scope is [16900 Ω, 144100 Ω], and result then is sent into signal turns
Mold changing block Simulink-Ps Convert, which are converted to, is input to variable resistor module after physical signalling, so as to realize that memristor hinders
The change of value；Wherein, ROFFResistance value during memristor whole undoped is represented, Δ R represents ROFF-RON；Current Sensor
It is respectively that power supply is perceived with voltage with Voltage Sensor, effect is that the electric current of correspondence position in measuring circuit and voltage are believed
Number.
Step 2：The signaling conversion circuit of cell neural network and memristor module is built, as shown in Fig. 2 specifically including：
Physical signalling is converted a signal into using Simulink-Ps convert, SimScape control is then inputted
Voltage source processed, you can to obtain SimScape voltage source pulses, be converted to Simulink voltage pulse signals so as to realize
The signal conversion of SimScape signals.
Because memristor model is built based on SimScape physical components, i.e., simulate memristor using variable resistor.Institute
When being integrated into the CNN models that Simulink is built, to carry out necessary signal conversion.Its function is to believe mathematical modeling
Number source is converted to SimScape signal sources.
Step 3：In Simulink, by the way that memristor first is built into wheatstone bridge configuration by way of parallel connection is connected again
Memristor Sudden-touch circuit.The present embodiment memristor is four, and memristor Sudden-touch circuit is by memristor M1 and M2, M3 and M4 difference antipoles
Property series connection, then by two tandem construction parallels, as shown in Figure 3.M1 to M4 resistance is respectively M1=M2=14.41k Ω, M2
=M3=1.69k Ω.
Step 4：It is integrated into after the memristor Sudden-touch circuit built using Simulink is encapsulated in cell neural network circuit,
A cell neural network circuit is built, as shown in figure 4, being specially：Using Simulink Subsystem by memristor bridge cynapse
Circuit package is into the subsystem of a Two-port netwerk, and a port is the input of voltage source, and one is after synaptic weight
Current source output.
The magnitude of voltage that can obtain each node by circuit partial pressure rule is as follows：
In above formula, M1,M2,M3And M4Memristor value of each memristor in t is represented respectively.Memristor and electricity
Resistance equally follows the circuit law of partial pressure.Memristor bridge cynapse output voltage VoutEqual to node A and B voltage difference, it is expressed as follows：
Above formula can be rewritten the equation for representing relation between synaptic weight and synaptic input, as follows.
vout=ψ vin
The representation of the ψ extremely memristor synaptic weights of the present invention in above formula.
How different using the memristor synaptic weight circuit acquisition of the present invention illustrated below in conjunction with simulation result
Synaptic weight.
(1) regulation synaptic weight is 1, because initial weight is set to 1 can obtain the cynapse power of needs without change
Value.As shown in figure 5, in Fig. 5 (a) M1 and M2 weights situations of change, that Fig. 5 (b) is represented is M3 and M4 weights situations of change, Fig. 5
(c) the memristor synaptic weight situation of change represented.
(2) regulation synaptic weight is 0.5, and it is 5s in the 1V cycles now to apply an amplitude to be, dutycycle is %10.57 electricity
Press pulse.That Fig. 6 (a) is represented is M1 and M2 weights situations of change, and that Fig. 6 (b) is represented is M3 and M4 weights situations of change, Fig. 6
(c) the memristor synaptic weight situation of change represented, it can be seen that weights stabilize to 0.5.
(3) regulation synaptic weight is 0, and it is 5s in the 1V cycles now to apply an amplitude to be, dutycycle is %18.60 voltage
Pulse.That Fig. 7 (a) is represented is M1 and M2 weights situations of change, and that Fig. 7 (b) is represented is M3 and M4 weights situations of change, Fig. 7 (c)
Memristor synaptic weight situation of change, it can be seen that weights stabilize to 0.
(4) regulation memristor synaptic weight is -1, and it is 5s in the 1V cycles now to apply an amplitude to be, dutycycle is %36.4's
Voltage pulse.That Fig. 8 (a) is represented is M1 and M2 weights situations of change, and that Fig. 8 (b) is represented is M3 and M4 weights situations of change, figure
The memristor synaptic weight situation of change of 8 (c), it can be seen that weights stabilize to -1.
Memristor synaptic weight circuit package is integrated into cell neural network circuit into subsystem, Fig. 9 and figure can be obtained
Synapse cell Weight template circuit shown in 10.The CNN feedback templates that Fig. 9 is represented, the CNN Control architectures that Figure 10 is represented, this is imitative
True explanation is 3 × 3 using template size
That (b) is represented in the Control architecture B that (a) is represented in adjusting template weights such as Figure 11, figure, figure is feedback template A.
The initial value figure of each cell of network when Figure 12 (a) is emulation, (b) is the output figure of network；The net that Figure 13 is represented
The last state of the cell of each in network.
When the use memristor cell neural network that Figure 14 is represented is emulated, cell C (2,2) state is with exporting with the time
Variation diagram, from the figure, it can be seen that network can illustrate memristor cynapse proposed by the present invention and biography to reach stabilization in 500us
The synapse cell of system can realize same function.
Above disclosed is only a kind of preferred embodiment of the invention, it is impossible to the right model of the present invention is limited with this
Enclose, therefore the equivalent variations made according to the claims in the present invention, still belong to the scope that the present invention is covered.
Claims (1)
1. a kind of adjustable weights modular simulation method of cell neural network, it is characterised in that comprise the following steps：
Step 1：The Simulink models of memristor are built, are specifically included：
1-1：The integration to memristor state variable w (t) is realized using Simulink integration module, wherein solver is used
ode45；
1-2：Memristor window function is realized using Simulink SQL module
Wherein, D is the thickness parameter of memristor, and window function parameter p is 10；
1-3：Realized using Simulink Gain modulesWherein, uvFor drift speed, RONRepresent that memristor is all mixed
Resistance value when miscellaneous；
1-4：Calculate ROFF- w (t) * Δ R, are then sent to signal conversion module Simulink-Ps Convert conversions by result
To be input to variable resistor module after physical signalling, so as to realize the change of memristor resistance；Wherein, ROFFRepresent that memristor is complete
Resistance value during portion's undoped, Δ R=ROFF-RON；
Step 2：The signaling conversion circuit of cell neural network and memristor module is built, is specifically included：
Physical signalling is converted a signal into using Simulink-Ps convert, SimScape control electricity is then inputted
Potential source, you can to obtain SimScape voltage source pulses, be converted to Simulink voltage pulse signals so as to realize
The signal conversion of SimScape signals；
Step 3：In Simulink, the memristor by the way that memristor first to be built to wheatstone bridge configuration by way of parallel connection is connected again
Sudden-touch circuit；
Step 4：It is integrated into after the memristor Sudden-touch circuit built using Simulink is encapsulated in cell neural network circuit, specifically
For：Memristor bridge Sudden-touch circuit is packaged into the subsystem of a Two-port netwerk using Simulink Subsystem, a port is
The input of voltage source, one is the current source output after synaptic weight.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710169625.4A CN107085628A (en) | 2017-03-21 | 2017-03-21 | A kind of adjustable weights modular simulation method of cell neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710169625.4A CN107085628A (en) | 2017-03-21 | 2017-03-21 | A kind of adjustable weights modular simulation method of cell neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107085628A true CN107085628A (en) | 2017-08-22 |
Family
ID=59614961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710169625.4A Pending CN107085628A (en) | 2017-03-21 | 2017-03-21 | A kind of adjustable weights modular simulation method of cell neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107085628A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108804786A (en) * | 2018-05-26 | 2018-11-13 | 江西理工大学 | A kind of memristor precircuit design method that associative neural network synaptic weight is plastic |
| CN109102072A (en) * | 2018-08-31 | 2018-12-28 | 江西理工大学 | A kind of memristor cynapse impulsive neural networks circuit design method based on single-electronic transistor |
| CN109816096A (en) * | 2019-01-23 | 2019-05-28 | 长安大学 | A kind of perceptron neural network circuit and its adjusting method based on memristor |
| WO2019127363A1 (en) * | 2017-12-29 | 2019-07-04 | 清华大学 | Weight coding method for neural network, computing apparatus, and hardware system |
| CN110738619A (en) * | 2019-10-15 | 2020-01-31 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
| CN113516138A (en) * | 2021-07-21 | 2021-10-19 | 中国民航大学 | Image processing method based on dual-mode memristor bridge synaptic circuit |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573238A (en) * | 2015-01-09 | 2015-04-29 | 江西理工大学 | Circuit design method for memory resisting cell neural network |
-
2017
- 2017-03-21 CN CN201710169625.4A patent/CN107085628A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573238A (en) * | 2015-01-09 | 2015-04-29 | 江西理工大学 | Circuit design method for memory resisting cell neural network |
Non-Patent Citations (4)
| Title |
|---|
| VALERI M. MLADENOV等: "MEMRISTOR MODELING IN MATLAB & PSPICE", 《PROCESSINGS 29TH EUROPEAN CONFERENCE ON MODELING AND SIMULATION》 * |
| 夏思为 等: "基于忆阻神经网络PID控制器设计", 《计算机学报》 * |
| 段美涛: "基于STDP规则的忆阻神经网络及应用", 《中国优秀硕士学位论文全文数据库工程科技II辑》 * |
| 胡小方: "基于忆阻器的非易失性存储器研究", 《中国优秀硕士学位论文全文数据库工程科技辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019127363A1 (en) * | 2017-12-29 | 2019-07-04 | 清华大学 | Weight coding method for neural network, computing apparatus, and hardware system |
| CN108804786A (en) * | 2018-05-26 | 2018-11-13 | 江西理工大学 | A kind of memristor precircuit design method that associative neural network synaptic weight is plastic |
| CN108804786B (en) * | 2018-05-26 | 2022-04-15 | 江西理工大学 | Memristive model circuit design method for realizing plasticity of synaptic weights of associative neural network |
| CN109102072A (en) * | 2018-08-31 | 2018-12-28 | 江西理工大学 | A kind of memristor cynapse impulsive neural networks circuit design method based on single-electronic transistor |
| CN109102072B (en) * | 2018-08-31 | 2021-11-23 | 江西理工大学 | Memristor synaptic pulse neural network circuit design method based on single-electron transistor |
| CN109816096A (en) * | 2019-01-23 | 2019-05-28 | 长安大学 | A kind of perceptron neural network circuit and its adjusting method based on memristor |
| CN109816096B (en) * | 2019-01-23 | 2022-10-18 | 长安大学 | Memristor-based perceptron neural network circuit and adjusting method thereof |
| CN110738619A (en) * | 2019-10-15 | 2020-01-31 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
| CN110738619B (en) * | 2019-10-15 | 2022-03-01 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
| CN113516138A (en) * | 2021-07-21 | 2021-10-19 | 中国民航大学 | Image processing method based on dual-mode memristor bridge synaptic circuit |
| CN113516138B (en) * | 2021-07-21 | 2022-07-29 | 中国民航大学 | Image processing method based on dual-mode memristor bridge synaptic circuit |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107085628A (en) | A kind of adjustable weights modular simulation method of cell neural network | |
| Li et al. | New synchronization criteria for memristor-based networks: adaptive control and feedback control schemes | |
| Gu et al. | Stability and synchronization for Riemann-Liouville fractional-order time-delayed inertial neural networks | |
| Gu et al. | Synchronization for fractional-order time-delayed memristor-based neural networks with parameter uncertainty | |
| CN109816096B (en) | Memristor-based perceptron neural network circuit and adjusting method thereof | |
| CN106446310A (en) | Transistor and system modeling methods based on artificial neural network | |
| CN104335224A (en) | Neuron circuit and method | |
| CN109352644B (en) | Mechanical arm movement rhythm control method based on CPG (coherent population) neuron network | |
| CN106682735A (en) | BP neural network algorithm based on PID adjustment | |
| CN103095191A (en) | Switch reluctance motor memory sensor model modeling method | |
| CN104766485A (en) | Traffic light optimization time distribution method based on improved fuzzy control | |
| CN208547942U (en) | Simulate vector-matrix multiplication operation circuit | |
| CN108710770A (en) | A kind of accurate cynapse method of adjustment towards the study of multiple-pulse Neural Network Supervised | |
| CN109217664B (en) | A kind of Fuzzy PI of boost breadboardin load unit | |
| CN103559541A (en) | Back propagation method for out-of-order data stream in big data | |
| CN110298435B (en) | Memristor type full-domain value BP neural network circuit | |
| Sira-Ramirez | Sliding mode-/spl Delta/modulation control of a" buck" converter | |
| CN117273103A (en) | Third-order memristor chaotic neuron circuit model and establishment method | |
| CN108170032A (en) | A kind of method for improving piezo-stack type driver positioning accuracy | |
| CN115987086A (en) | Single-switch DC-DC converter on-line control method based on neural network | |
| CN109565246A (en) | For controlling the module of the internal energy of converter | |
| Zhu et al. | Memristor-based neural network PID controller for buck converter | |
| CN107038288A (en) | Power electronic equipment parallel equivalent expansion method and module based on Bei Jielong models | |
| Jiang et al. | Boundedness and stability for nonautonomous bidirectional associative neural networks with delay | |
| CN105388757A (en) | Compound control method for electric loading system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170822 |
|
| RJ01 | Rejection of invention patent application after publication |