CN109977534A - Circuit parameter optimization method and system based on intensified learning - Google Patents
Circuit parameter optimization method and system based on intensified learning Download PDFInfo
- Publication number
- CN109977534A CN109977534A CN201910221710.XA CN201910221710A CN109977534A CN 109977534 A CN109977534 A CN 109977534A CN 201910221710 A CN201910221710 A CN 201910221710A CN 109977534 A CN109977534 A CN 109977534A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- circuit
- optimal parameters
- parameter
- 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]
- G06F30/30—Circuit design
- G06F30/39—Circuit design at the physical level
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a kind of circuit parameter optimization method and system based on intensified learning, wherein this method comprises: obtaining the Optimal Parameters and observation information of artificial circuit, and Optimal Parameters are initialized；Observation information is inputted into trained neural network model in advance, to export the renewal amount of Optimal Parameters；Optimal Parameters are updated to reach optimization aim according to renewal amount.This method can quickly obtain Optimal Parameters optimal under given design parameter and design object, improve circuit layout efficiency, shorten the Time To Market of circuit product.
Description
Technical field
The present invention relates to IC design technical field, in particular to a kind of circuit parameter optimization based on intensified learning
Method and system.
Background technique
In circuit design work, parameter optimization is an inevitable problem, especially Analog Circuit Design, to setting
The skill requirement of meter person is higher.With the continuous improvement of footprint and performance, the demand of optimized increasingly increases
Greatly.Traditional optimization algorithm needs to carry out a large amount of iteration, due to needing that emulator is called to carry out calculating target function in iteration, adjusts
It is very big with the cost of emulator, so the optimization to circuit parameter requires a great deal of time.This greatly constrains circuit
The efficiency of design, not only increases the cost of labor of circuit design, and extends the Time To Market of circuit product, and then makes to produce
The competitiveness of product reduces.
In traditional parameter optimisation procedure, the experience for commonly relying on designer provides one preferably as a result, sometimes
Also optimal solution can be found by optimization algorithm.Traditional optimization algorithm include particle swarm optimization algorithm, differential evolution algorithm and
Genetic algorithm etc., these optimization algorithms are all based on search wide and randomness as far as possible to obtain globally optimal solution, need to consume
Take longer time, and under different circuit parameters and design object, needs to redesign and optimize.Secondly, these are calculated
Method is obtained by engineer, and to algorithm designer, more stringent requirements are proposed for this: designer should have problem sufficiently
Understanding, and have certain experience.
For the same circuit topology, just there is different optimal solutions under different design objects.It is more in number of parameters
In the case where, it is desirable that it with the target for meeting current circuit design is relatively difficult that new hand designer, which can obtain rapidly preferably parameter,
Thing.In addition, be frequently run onto circuit design process migration the problem of, from large scale technogenic migration to small size technique it
Afterwards, although circuit topology is constant, circuit parameter therein is required to designer and redesigns to meet design requirement.
Optimization problem constant for this structure, only being changed by parameter come control problem, proposes introducing intensified learning
Carry out a kind of " study " optimization algorithm, a certain topological circuit can be optimized according to optimization aim, and tied rapidly
Fruit.,
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, an object of the present invention is to provide a kind of circuit parameter optimization method based on intensified learning, the party
Method realizes the circuit emulation method that parallelization calculates by temporal partitioning so that transient analysis may be implemented it is efficient parallel
Change, improves the efficiency of the transient analysis of circuit, to improve circuit layout efficiency, shorten the Time To Market of circuit product.
It is another object of the present invention to propose a kind of circuit parameter optimization system based on intensified learning.
In order to achieve the above objectives, one aspect of the present invention embodiment proposes a kind of circuit parameter optimization based on intensified learning
Method, comprising: S1 obtains the Optimal Parameters and observation information of artificial circuit, and initializes to the Optimal Parameters；S2,
The observation information is inputted into trained neural network model in advance, to export the renewal amount of the Optimal Parameters；S3, according to
The renewal amount is updated to reach optimization aim the Optimal Parameters.
The circuit parameter optimization method based on intensified learning of the embodiment of the present invention is realized by intensified learning and is instructed under line
The circuit parameter optimization method practiced, tested on line, gives the input file of a circuit simulation, and this document is imitative for describing to need
Topological structure, external drive signal, Optimal Parameters and the design parameter of genuine circuit；Nitrification enhancement is utilized under online
Optimal circuit parameter path optimizing is practised, is used when then optimizing on line, quickly to obtain given design parameter and design
Optimal Optimal Parameters under target improve circuit layout efficiency, shorten the Time To Market of circuit product.
In addition, the circuit parameter optimization method according to the above embodiment of the present invention based on intensified learning can also have with
Under additional technical characteristic:
Further, in one embodiment of the invention, before S1, further includes:
Construct neural network model, wherein be trained using nitrification enhancement to neural network model, neural network
The input of model is the observation information in many kinds of parameters, loss function size and artificial circuit in artificial circuit, output
For the renewal amount of the Optimal Parameters of artificial circuit.
Further, in one embodiment of the invention, the building neural network model specifically includes:
A obtains and fixes one group of design parameter, initializes the Optimal Parameters；
The observation information of artificial circuit is inputted the neural network model put up, exports the Optimal Parameters by b
Renewal amount, and new loss function is calculated with the function that is recompensed；
The observation information, the renewal amount and the Reward Program composition triple are acted on intensified learning and calculated by c
Method, to update the parameter of neural network model；
D, iterative step b and step c, reaching the first preset condition terminates this training；
E repeats step a to step d until neural network model restrains to obtain the trained neural network mould in advance
Type.
Further, in one embodiment of the invention, further includes: to the trained neural network model in advance
It is tested；
Wherein, the specific steps of test are as follows:
F obtains the design parameter, initializes the Optimal Parameters；
G inputs the observation information to neural network model, updates the Optimal Parameters after exporting renewal amount；
H repeats step g, until reaching the second preset condition, terminates the test to neural network model.
Further, in one embodiment of the invention, the design parameter includes: supply voltage, process, swashs
Encourage signal and design object；The Optimal Parameters include: channel width and bias voltage.
In order to achieve the above objectives, it is excellent to propose a kind of circuit parameter based on intensified learning for another aspect of the present invention embodiment
Change system, comprising: processing module, for obtaining the Optimal Parameters and observation information of artificial circuit, and to the Optimal Parameters into
Row initialization；Output module, it is described excellent to export for the observation information to be inputted trained neural network model in advance
Change the renewal amount of parameter；Optimization module, for being updated to the Optimal Parameters according to the renewal amount to reach optimization mesh
Mark.
The circuit parameter optimization system based on intensified learning of the embodiment of the present invention is realized by intensified learning and is instructed under line
The circuit parameter optimization method practiced, tested on line, gives the input file of a circuit simulation, and this document is imitative for describing to need
Topological structure, external drive signal, Optimal Parameters and the design parameter of genuine circuit；Nitrification enhancement is utilized under online
Optimal circuit parameter path optimizing is practised, is used when then optimizing on line, quickly to obtain given design parameter and design
Optimal Optimal Parameters under target improve circuit layout efficiency, shorten the Time To Market of circuit product.
In addition, the circuit parameter optimization system according to the above embodiment of the present invention based on intensified learning can also have with
Under additional technical characteristic:
Further, in one embodiment of the invention, further includes: building module；
The building module instructs neural network model using nitrification enhancement for constructing neural network model
Practice, the input of neural network model is the sight in many kinds of parameters, loss function size and artificial circuit in artificial circuit
Measurement information exports the renewal amount of the Optimal Parameters for artificial circuit.
Further, in one embodiment of the invention, the building module is specifically used for,
One group of design parameter is obtained and fixed, the Optimal Parameters are initialized；
The observation information of artificial circuit is inputted to the neural network model put up, exports the Optimal Parameters more
New amount, and new loss function is calculated with the function that is recompensed；
The observation information, the renewal amount and the Reward Program composition triple are acted on into nitrification enhancement,
To update the parameter of neural network model；
It is iterated to update the parameter of neural network model, reaching the first preset condition terminates primary training；
Repeatedly training is carried out until neural network model restrains to obtain the trained neural network model in advance.
Further, in one embodiment of the invention, further includes: test module；
The test module is used to test the trained neural network model in advance.
Further, in one embodiment of the invention, the design parameter includes: supply voltage, process, swashs
Encourage signal and design object；The Optimal Parameters include: channel width and bias voltage.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, in which:
Fig. 1 is the circuit parameter optimization method flow chart based on intensified learning according to one embodiment of the invention；
Fig. 2 is the schematic diagram according to a circuit input file example of one embodiment of the invention；
Fig. 3 is the schematic diagram according to the training neural network model of one embodiment of the invention；
Fig. 4 is according to the experimental results of one embodiment of the invention and compared with the optimum results of differential evolution algorithm
Figure；
Fig. 5 is the circuit parameter optimization system structural schematic diagram based on intensified learning according to one embodiment of the invention.
Specific embodiment
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.
The circuit parameter optimization method based on intensified learning proposed according to embodiments of the present invention is described with reference to the accompanying drawings
And system.
The circuit parameter optimization side based on intensified learning proposed according to embodiments of the present invention is described with reference to the accompanying drawings first
Method.
Fig. 1 is the circuit parameter optimization method flow chart based on intensified learning according to one embodiment of the invention.
As shown in Figure 1, should circuit parameter optimization method based on intensified learning the following steps are included:
In step sl, the Optimal Parameters and observation information of artificial circuit are obtained, and Optimal Parameters are initialized.
It further, should as shown in Fig. 2, first giving the input file of a circuit simulation before circuit parameter optimization
File is used for topological structure, external drive signal, Optimal Parameters and the design parameter for describing to need the circuit emulated, wherein
Design parameter includes supply voltage, process, pumping signal and design object etc., and Optimal Parameters include channel width and biasing
Voltage etc., as soon as it there is known given group design parameter to circuit multiple parameters and design object, according to design
Target adjustment Optimal Parameters, to reach design object.
It should be noted that the loss function of design object is also defined during optimizing, wherein loss
It is better that the smaller then expression parameter of function selects.
In step s 2, observation information is inputted into trained neural network model in advance, to export Optimal Parameters more
New amount.
Further, a neural network model is first constructed before step S1, for exporting the renewal amount of Optimal Parameters.
Neural network model is trained neural network model using nitrification enhancement, the input of neural network model
For the observation information in many kinds of parameters, loss function size and the artificial circuit in artificial circuit, export as the excellent of artificial circuit
Change the renewal amount of parameter.
Further, as shown in figure 3, building neural network model specifically includes:
A obtains and fixes one group of design parameter, initializes Optimal Parameters；
The observation information of artificial circuit is inputted the neural network model put up, exports the renewal amount of Optimal Parameters by b,
And new loss function is calculated with the function that is recompensed；
Observation information, renewal amount and Reward Program composition triple are acted on nitrification enhancement, to update nerve by c
The parameter of network model；
D, iterative step b and step c, reaching the first preset condition terminates this training；
E repeats step a to step d until neural network model restrains to obtain preparatory trained neural network model.
Specifically, according to the observation information of one group of design parameter, Optimal Parameters and artificial circuit, to the depth tentatively built
Neural network model is trained to obtain the renewal amount of Optimal Parameters, calculates new loss function, takes negative to loss function
Be recompensed function, by observation information, renewal amount and Reward Program composition triple to nitrification enhancement, Lai Gengxin depth mind
Parameter through network model reaches maximum train epochs or other conditions after parameter repeatedly updates, then this training knot
Beam repeats circulation above-mentioned steps, and until trained deep neural network model restrains, deep neural network model has been trained
At.
It should be noted that can also be tested model after by deep neural network model training.
Further, in one embodiment of the invention, the specific steps of test are as follows:
F obtains design parameter, initializes Optimal Parameters；
G inputs observation information to neural network model, updates Optimal Parameters after exporting renewal amount；
H repeats step g, until reaching the second preset condition, terminates the test to neural network model.
Specifically, observation information is inputted depth nerve to Optimal Parameters initialization by designer's given design parameter first
In network model, output renewal amount is updated Optimal Parameters, is repeatedly updated to Optimal Parameters, reaches maximum excellent
After changing step number or obtaining satisfied Optimal Parameters, finish test procedure, as shown in figure 4, illustrating the knot of the test of this model
Fruit and the result being compared with the optimum results of differential evolution algorithm (Difference Evolution, DE).
In step s3, Optimal Parameters are updated to reach optimization aim according to renewal amount.
Specifically, Optimal Parameters are updated according to the renewal amount that deep neural network model exports, reach design ginseng
Design object in number realizes the parameter optimization to circuit.
In an embodiment of the present invention, include optimization algorithm, initialize optimized variable first, calculate the update currently walked
Amount updates optimized variable.This set of optimization algorithm frame is widely used in optimization algorithm structure of today, different optimization algorithms
It is only different to the calculation formula of renewal amount.
The circuit parameter optimization method specific steps based on intensified learning are described below, as follows:
(1) give a circuit simulation input file, this document be used for describe needs emulate circuit topological structure,
External drive signal, Optimal Parameters and design parameter, design parameter include supply voltage, process, pumping signal and set
Target etc. is counted, Optimal Parameters then include channel width, bias voltage etc., in the case where giving one group of design parameter, are adjusted excellent
Change parameter to reach optimization aim；
(2) loss function of design object is defined, the smaller expression parameter of loss function selects better；
(3) neural network is built, network is trained using nitrification enhancement, the input of neural network is in circuit
Parameter, other observation informations in loss function size and circuit, export as the renewal amount of circuit optimization parameter；
(4) step of single training are as follows:
1. one group of design parameter is randomly choosed in design parameter given range, it is constant before this training terminates；
2. initializing Optimal Parameters；
3. optimized variable is updated after output renewal amount, calculates new loss function to neural network inputs observation information,
Reward Program is defined as the negative of loss function；
4. by the triple of observation information, renewal amount and Reward Program composition to nitrification enhancement, for updating nerve
Network parameter；
5. circulation carry out 3.~4. step terminate this instruction until reaching maximum train epochs or other conditions
Practice；
(5) constantly circulation carries out step (4), until training pattern restrains；
(6) trained model is tested, testing procedure are as follows:
1. designer's given design parameter；
2. initializing Optimal Parameters；
3. updating optimized variable after output renewal amount to neural network inputs observation information；
4. carrying out multiple 3. step, after reaching largest optimization step number or obtaining satisfied Optimal Parameters, terminate to survey
Examination process.
The circuit parameter optimization method based on intensified learning proposed according to embodiments of the present invention, is realized by intensified learning
The circuit parameter optimization method trained under line, tested on line gives the input file of a circuit simulation, and this document is for describing
Topological structure, external drive signal, Optimal Parameters and the design parameter for the circuit for needing to emulate；Intensified learning is utilized under online
Algorithm learns to optimal circuit parameter path optimizing, uses when then optimizing on line, quickly to obtain given design parameter
And optimal Optimal Parameters under design object, circuit layout efficiency is improved, the Time To Market of circuit product is shortened.
The circuit parameter optimization system based on intensified learning proposed according to embodiments of the present invention is described referring next to attached drawing.
Fig. 5 is the circuit parameter optimization system structural schematic diagram based on intensified learning according to one embodiment of the invention.
As shown in figure 5, being somebody's turn to do the circuit parameter optimization system 10 based on intensified learning includes: processing module 100, output module
200 and optimization module 300.
Wherein, processing module 100 is used to obtain the Optimal Parameters and observation information of artificial circuit, and carries out to Optimal Parameters
Initialization.Output module 200 is used to inputting observation information into trained neural network model in advance, to export Optimal Parameters
Renewal amount.Optimization module 300 is for being updated to reach optimization aim Optimal Parameters according to renewal amount.The circuit parameter is excellent
Change system 10 can quickly obtain Optimal Parameters optimal under given design parameter and design object, improve circuit layout efficiency,
Shorten the Time To Market of circuit product.
Further, in one embodiment of the invention, further includes: building module；
Building module is used to construct neural network model, is trained using nitrification enhancement to neural network model,
The input of neural network model is the observation information in many kinds of parameters, loss function size and artificial circuit in artificial circuit,
Output is the renewal amount of the Optimal Parameters of artificial circuit.
Further, in one embodiment of the invention, building module is specifically used for,
One group of design parameter is obtained and fixed, Optimal Parameters are initialized；
The observation information of artificial circuit is inputted to the neural network model put up, exports the renewal amount of Optimal Parameters, and
New loss function is calculated with the function that is recompensed；
Observation information, renewal amount and Reward Program composition triple are acted on into nitrification enhancement, to update nerve net
The parameter of network model；
It is iterated to update the parameter of neural network model, reaching the first preset condition terminates primary training；
Repeatedly training is carried out until neural network model restrains to obtain preparatory trained neural network model.
Further, in one embodiment of the invention, further includes: test module；
Test module is used to test preparatory trained neural network model.
Further, in one embodiment of the invention, design parameter includes: supply voltage, process, excitation letter
Number and design object；Optimal Parameters include: channel width and bias voltage.
It should be noted that the aforementioned explanation to the circuit parameter optimization method embodiment based on intensified learning is also fitted
For the system of the embodiment, details are not described herein again.
The circuit parameter optimization system based on intensified learning proposed according to embodiments of the present invention, is realized by intensified learning
The circuit parameter optimization method trained under line, tested on line gives the input file of a circuit simulation, and this document is for describing
Topological structure, external drive signal, Optimal Parameters and the design parameter for the circuit for needing to emulate；Intensified learning is utilized under online
Algorithm learns to optimal circuit parameter path optimizing, uses when then optimizing on line, quickly to obtain given design parameter
And optimal Optimal Parameters under design object, circuit layout efficiency is improved, the Time To Market of circuit product is shortened.
In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of circuit parameter optimization method based on intensified learning, which comprises the following steps:
S1 obtains the Optimal Parameters and observation information of artificial circuit, and initializes to the Optimal Parameters；
The observation information is inputted trained neural network model in advance, to export the renewal amount of the Optimal Parameters by S2；
S3 is updated to reach optimization aim the Optimal Parameters according to the renewal amount.
2. the circuit parameter optimization method according to claim 1 based on intensified learning, which is characterized in that before S1,
Further include:
Construct neural network model, wherein be trained using nitrification enhancement to neural network model, neural network model
Input be many kinds of parameters in artificial circuit, the observation information in loss function size and artificial circuit, it is imitative for exporting
The renewal amount of the Optimal Parameters of true circuit.
3. the circuit parameter optimization method according to claim 2 based on intensified learning, which is characterized in that the building mind
It is specifically included through network model:
A obtains and fixes one group of design parameter, initializes the Optimal Parameters；
The observation information of artificial circuit is inputted the neural network model put up, exports the update of the Optimal Parameters by b
Amount, and new loss function is calculated with the function that is recompensed；
The observation information, the renewal amount and the Reward Program composition triple are acted on nitrification enhancement by c, with
Update the parameter of neural network model；
D, iterative step b and step c, reaching the first preset condition terminates this training；
E repeats step a to step d until neural network model restrains to obtain the trained neural network model in advance.
4. the circuit parameter optimization method according to claim 3 based on intensified learning, which is characterized in that further include: it is right
The trained neural network model in advance is tested；
Wherein, the specific steps of test are as follows:
F obtains the design parameter, initializes the Optimal Parameters；
G inputs the observation information to neural network model, updates the Optimal Parameters after exporting renewal amount；
H repeats step g, until reaching the second preset condition, terminates the test to neural network model.
5. the circuit parameter optimization method according to claim 3 based on intensified learning, which is characterized in that
The design parameter includes: supply voltage, process, pumping signal and design object；
The Optimal Parameters include: channel width and bias voltage.
6. a kind of circuit parameter optimization system based on intensified learning characterized by comprising
Processing module is initialized for obtaining the Optimal Parameters and observation information of artificial circuit, and to the Optimal Parameters；
Output module, for the observation information to be inputted trained neural network model in advance, to export the optimization ginseng
Several renewal amounts；
Optimization module, for being updated the Optimal Parameters to reach optimization aim according to the renewal amount.
7. the circuit parameter optimization system according to claim 6 based on intensified learning, which is characterized in that further include: structure
Model block；
The building module is used to construct neural network model, is trained using nitrification enhancement to neural network model,
The input of neural network model is the observation letter in many kinds of parameters, loss function size and artificial circuit in artificial circuit
Breath, exports the renewal amount of the Optimal Parameters for artificial circuit.
8. the circuit parameter optimization system according to claim 7 based on intensified learning, which is characterized in that the building mould
Block is specifically used for,
One group of design parameter is obtained and fixed, the Optimal Parameters are initialized；
The observation information of artificial circuit is inputted to the neural network model put up, exports the update of the Optimal Parameters
Amount, and new loss function is calculated with the function that is recompensed；
The observation information, the renewal amount and the Reward Program composition triple are acted on into nitrification enhancement, with more
The parameter of new neural network model；
It is iterated to update the parameter of neural network model, reaching the first preset condition terminates primary training；
Repeatedly training is carried out until neural network model restrains to obtain the trained neural network model in advance.
9. the circuit parameter optimization system according to claim 6 based on intensified learning, which is characterized in that further include: it surveys
Die trial block；
The test module is used to test the trained neural network model in advance.
10. the circuit parameter optimization system according to claim 8 based on intensified learning, which is characterized in that
The design parameter includes: supply voltage, process, pumping signal and design object；
The Optimal Parameters include: channel width and bias voltage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910221710.XA CN109977534A (en) | 2019-03-22 | 2019-03-22 | Circuit parameter optimization method and system based on intensified learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910221710.XA CN109977534A (en) | 2019-03-22 | 2019-03-22 | Circuit parameter optimization method and system based on intensified learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109977534A true CN109977534A (en) | 2019-07-05 |
Family
ID=67080038
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910221710.XA Pending CN109977534A (en) | 2019-03-22 | 2019-03-22 | Circuit parameter optimization method and system based on intensified learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109977534A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112417803A (en) * | 2020-12-02 | 2021-02-26 | 苏州复鹄电子科技有限公司 | Artificial intelligence algorithm-based automatic optimization scheme for design parameters of analog integrated circuit |
| CN112883672A (en) * | 2021-03-05 | 2021-06-01 | 清华大学 | Digital modeling method and device for mixed signal circuit |
| CN113158608A (en) * | 2021-02-26 | 2021-07-23 | 北京大学 | Processing method, device and equipment for determining parameters of analog circuit and storage medium |
| CN113204539A (en) * | 2021-05-12 | 2021-08-03 | 南京大学 | Big data system parameter automatic optimization method fusing system semantics |
| CN113239651A (en) * | 2021-07-12 | 2021-08-10 | 苏州贝克微电子有限公司 | Artificial intelligence implementation method and system for circuit design |
| CN113283213A (en) * | 2021-07-19 | 2021-08-20 | 苏州贝克微电子有限公司 | Circuit design system and method based on machine learning |
| CN113609802A (en) * | 2020-05-04 | 2021-11-05 | 辉达公司 | Routing connections in reinforcement-based integrated circuits |
| CN114139472A (en) * | 2021-11-04 | 2022-03-04 | 江阴市智行工控科技有限公司 | Integrated circuit direct current analysis method and system based on reinforcement learning dual-model structure |
| CN114611449A (en) * | 2022-04-25 | 2022-06-10 | 杭州法动科技有限公司 | Rapid simulation optimization method and system for analog circuit |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107272785A (en) * | 2017-07-19 | 2017-10-20 | 北京上格云技术有限公司 | A kind of electromechanical equipment and its control method, computer-readable medium |
| CN109343341A (en) * | 2018-11-21 | 2019-02-15 | 北京航天自动控制研究所 | It is a kind of based on deeply study carrier rocket vertically recycle intelligent control method |
-
2019
- 2019-03-22 CN CN201910221710.XA patent/CN109977534A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107272785A (en) * | 2017-07-19 | 2017-10-20 | 北京上格云技术有限公司 | A kind of electromechanical equipment and its control method, computer-readable medium |
| CN109343341A (en) * | 2018-11-21 | 2019-02-15 | 北京航天自动控制研究所 | It is a kind of based on deeply study carrier rocket vertically recycle intelligent control method |
Non-Patent Citations (2)
| Title |
|---|
| CHANGCHENG TANG 等: "Parametric Circuit Optimization with Reinforcement Learning", 《2018 IEEE COMPUTER SOCIETY ANNUAL SYMPOSIUM ON VLSI (ISVLSI)》 * |
| 唐长成 等: "基于强化学习的参数化电路优化算法", 《微电子学与计算机》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113609802A (en) * | 2020-05-04 | 2021-11-05 | 辉达公司 | Routing connections in reinforcement-based integrated circuits |
| CN112417803B (en) * | 2020-12-02 | 2024-02-06 | 苏州复鹄电子科技有限公司 | Automatic optimization method for design parameters of analog integrated circuit based on artificial intelligence algorithm |
| CN112417803A (en) * | 2020-12-02 | 2021-02-26 | 苏州复鹄电子科技有限公司 | Artificial intelligence algorithm-based automatic optimization scheme for design parameters of analog integrated circuit |
| CN113158608A (en) * | 2021-02-26 | 2021-07-23 | 北京大学 | Processing method, device and equipment for determining parameters of analog circuit and storage medium |
| CN112883672A (en) * | 2021-03-05 | 2021-06-01 | 清华大学 | Digital modeling method and device for mixed signal circuit |
| CN113204539A (en) * | 2021-05-12 | 2021-08-03 | 南京大学 | Big data system parameter automatic optimization method fusing system semantics |
| CN113204539B (en) * | 2021-05-12 | 2023-08-22 | 南京大学 | Automatic optimization method for big data system parameters fusing system semantics |
| CN113239651B (en) * | 2021-07-12 | 2021-09-17 | 苏州贝克微电子有限公司 | Artificial intelligence implementation method and system for circuit design |
| CN113239651A (en) * | 2021-07-12 | 2021-08-10 | 苏州贝克微电子有限公司 | Artificial intelligence implementation method and system for circuit design |
| CN113283213B (en) * | 2021-07-19 | 2021-11-09 | 苏州贝克微电子有限公司 | Circuit design system and method based on machine learning |
| CN113283213A (en) * | 2021-07-19 | 2021-08-20 | 苏州贝克微电子有限公司 | Circuit design system and method based on machine learning |
| CN114139472A (en) * | 2021-11-04 | 2022-03-04 | 江阴市智行工控科技有限公司 | Integrated circuit direct current analysis method and system based on reinforcement learning dual-model structure |
| CN114139472B (en) * | 2021-11-04 | 2023-05-02 | 江阴市智行工控科技有限公司 | Integrated circuit direct current analysis method and system based on reinforcement learning dual-mode structure |
| CN114611449A (en) * | 2022-04-25 | 2022-06-10 | 杭州法动科技有限公司 | Rapid simulation optimization method and system for analog circuit |
| CN114611449B (en) * | 2022-04-25 | 2024-07-30 | 杭州法动科技有限公司 | Quick simulation optimization method and system for analog circuit |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109977534A (en) | Circuit parameter optimization method and system based on intensified learning | |
| CN105260471B (en) | Commodity personalized ordering model training method and system | |
| CN108875161A (en) | Flow grade prediction technique based on convolutional neural networks deep learning | |
| CN106779068A (en) | The method and apparatus for adjusting artificial neural network | |
| CN108133258A (en) | A kind of mixing global optimization method | |
| CN111045326A (en) | Tobacco shred drying process moisture prediction control method and system based on recurrent neural network | |
| CN106874688A (en) | Intelligent lead compound based on convolutional neural networks finds method | |
| CN109376933A (en) | Lithium ion battery negative material energy density prediction technique neural network based | |
| AT512977A2 (en) | Method for determining a model of an output of a technical system | |
| CN109325613B (en) | Method for online predicting static voltage stability margin by using parallel sequential extreme learning machine | |
| CN105469142A (en) | Neural network increment-type feedforward algorithm based on sample increment driving | |
| CN105184400A (en) | Tobacco field soil moisture prediction method | |
| Pazouki | A practical surface irrigation system design based on volume balance model and multi-objective evolutionary optimization algorithms | |
| CN108142976B (en) | Cut tobacco drying process parameter optimization method | |
| CN106842948A (en) | The method for optimally controlling of the HDP molecular distillation systems based on BP networks | |
| CN115964934A (en) | Artificial intelligence cross-platform model intelligent calculation engine construction method based on knowledge self-evolution | |
| CN106355273A (en) | Predication system and predication method for after-stretching performance of nuclear material radiation based on extreme learning machine | |
| CN104573216A (en) | Antenna performance optimizing method and device | |
| EP3682301B1 (en) | Randomized reinforcement learning for control of complex systems | |
| CN110851920A (en) | Automatic generation method for main rib line of die pressing device | |
| CN109635932A (en) | A kind of Graphic Design and implementation method of part connection multilayer perceptron | |
| KR20230090915A (en) | Method for closed-loop linear model gain closed-loop update using artificial neural network | |
| CN113569898A (en) | Water quality model parameter inversion analysis method and system based on BP neural network | |
| Sugunnasil et al. | Modelling a neural network using an algebraic method | |
| CN110490307A (en) | A method of beta pruning optimization system in channel is carried out based on network model |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190705 |