CN109656721A - A kind of efficient intelligence system - Google Patents
A kind of efficient intelligence system Download PDFInfo
- Publication number
- CN109656721A CN109656721A CN201811628648.8A CN201811628648A CN109656721A CN 109656721 A CN109656721 A CN 109656721A CN 201811628648 A CN201811628648 A CN 201811628648A CN 109656721 A CN109656721 A CN 109656721A
- Authority
- CN
- China
- Prior art keywords
- artificial intelligence
- controller
- intelligence task
- module
- processing module
- 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
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 50
- 238000000034 method Methods 0.000 claims description 18
- 238000003062 neural network model Methods 0.000 claims description 9
- 238000003672 processing method Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 description 3
- 238000007906 compression Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000153 supplemental effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Programmable Controllers (AREA)
Abstract
The invention discloses a kind of artificial intelligence task processing systems, belong to artificial intelligence field, system includes: memory module and logic processing module, wherein memory module further includes controller and storage unit, the controller is when logic processing module is in non-working condition, artificial intelligence task is handled, and exports multiple processing results；Further, in logic processing module working condition, multiple output results is further processed, final result is obtained and exports.When executing intelligent algorithm the runing time for executing intelligent algorithm, the resource of system when improving the execution efficiency of algorithm, while taking full advantage of idle state can be greatly shortened under the premise of guaranteeing intelligent algorithm running precision in the present invention.
Description
Technical field
The present invention relates to an artificial intelligence field more particularly to a kind of artificial intelligence task processing systems.
Background technique
With the development of intelligent algorithm especially machine deep learning, more and more to apply in visual identity, voice is known
Not, extensive development has been obtained on picture analyzing, by taking deep learning as an example, with the increasing of neuronal quantity in deep learning model
Add, the prediction accuracy of model can also increase therewith, but need the time of operation that can also increase therewith, cause the damage of performance
It loses.Intelligent algorithm especially neural network deep learning obtains algorithm and generally requires to handle very big data simultaneously, is handling
Complicated calculation and a large amount of data of operation are handled in the limited situation of resource and generally requires many time, cause excessive resource
And power consumption.
Summary of the invention
According to defect of the existing technology, a kind of efficient intelligent artificial intelligence task processing system is now provided,
System leisure state is pre-processed, and is specifically included:
A kind of artificial intelligence task processing system, wherein including memory module and logic processing module, the logical process
Module is connected with the storage storing module；
The memory module includes storage unit and the controller for connecting the storage unit, is saved in the storage unit
Initial data in relation to being coupled to artificial intelligence task；
The neural network model for being handled the initial data is preset in the controller, when the logic
When processing module is not in working condition, the controller using the neural network model to the initial data at
Reason, and export multiple processing results；
When the logic processing module is in running order, the logic processing module is exported according to the controller
Multiple processing results are further processed, to obtain the final process result for being associated with the artificial intelligence task and defeated
Out.
Preferably, in the artificial intelligence task processing system, the logic processing module ties multiple processing
The mode that fruit is further processed are as follows: screening obtains accuracy highest one and as institute in multiple processing results
State final process result output.
Preferably, in the artificial intelligence task processing system, the preset neural network mould in the controller
Type is the neural network model through overcompression.
Preferably, in the artificial intelligence task processing system, the initial data that is saved in the storage unit
For the initial data through overcompression.
Preferably, in the artificial intelligence task processing system, the memory module is nonvolatile memory.
Preferably, in the artificial intelligence task processing system, the controller is microcontroller or microprocessor.
Preferably, in the artificial intelligence task processing system, in Yu Suoshu controller, the described of output is preset
The quantity of processing result.
Preferably, a kind of artificial intelligence task processing method is also wrapped applied to the artificial intelligence task processing system
It includes:
When step S1, Yu Suoshu logic processing module is not at working condition, the controller is in the storage unit
The initial data for being associated with the artificial intelligence task saved is handled, and exports multiple processing results；
Step S2, when the logic processing module is in running order, the logic processing module is according to the control
Multiple processing results of device output are further processed, to obtain being associated with the final process of the artificial intelligence task
As a result it and exports.
Preferably, the artificial intelligence task processing method, wherein in the step S2, the logic processing module
The mode that multiple processing results are further processed are as follows: screening obtains accuracy most in multiple processing results
It high one and is exported as the final process result.
The beneficial effect of above-mentioned technical proposal is:
It, can be with when executing intelligent algorithm by pre-processing intelligent algorithm and data in system leisure state
Under the premise of guaranteeing intelligent algorithm running precision, the runing time for executing intelligent algorithm is greatly shortened, is promoted
The execution efficiency of algorithm, the resource of system when taking full advantage of idle state ensure that the working efficiency of artificial intelligence system,
Also save resource.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of artificial intelligence system in present pre-ferred embodiments；
Fig. 2 is in present pre-ferred embodiments, and controller pre-processes artificial intelligence task schematic diagram；
Fig. 3 is artificial intelligence system work step flow diagram in present pre-ferred embodiments.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art without creative labor it is obtained it is all its
His embodiment, shall fall within the protection scope of the present invention.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase
Mutually combination.
The present invention will be further explained below with reference to the attached drawings and specific examples, but not as the limitation of the invention.
Existing intelligent algorithm, the neuronal quantity in the accuracy and learning model of intelligent algorithm operation is at just
Than, i.e., if it is desired to obtaining high-precision operational model, the quantity of neuron in learning model must be just improved, still, this companion
Be to need largely to be calculated, in the prior art, can only under limited resources, by carry out a large amount of data operation with
Complicated algorithm process, in this manner, artificial intelligence system learning time is long and easily causes the wasting of resources.
To solve the above problems, providing one kind can be carried out pretreated artificial intelligence task in present pre-ferred embodiments
Processing system, as shown in Figure 1, comprising:
Memory module 1 and logic processing module 2, logic processing module 2 are connected with storage module 1, and wherein memory module 1 is wrapped
Storage unit 11 and controller 12 are included, wherein when logic processing module 2 is in standby or off position, by storing mould
Controller 12 in block 1 pre-processes the artificial intelligence task of logic processing module by intelligent algorithm；
Further, it is preset in present pre-ferred embodiments, in above controller for being carried out to the initial data
The neural network model of processing, when the logic processing module is not in working condition, the controller uses the nerve
Network model handles the initial data, and exports multiple processing results.
Further, in present pre-ferred embodiments, when the logic processing module is in running order, the logic
Processing module is further processed according to multiple processing results that the controller exports, to obtain being associated with the people
The final process result of work intelligent task and output.
In present pre-ferred embodiments, above-mentioned memory module 1 includes storage unit 11 and controller 12, wherein storage unit
11 be nonvolatile memory, and controller 12 can be microcontroller or microprocessor.
In present pre-ferred embodiments, above-mentioned logic processing module 2, including but not limited to central processing unit (CPU), figure
As processing unit (GPU), tensor processing unit (TPU), field programmable gate array (FPGA), specific integrated circuit (ASIC).
In conclusion in present pre-ferred embodiments, when system works can first the working condition to logic processing module 2 into
Row judgement, when logic processing module 2 is in standby or off position, controller 12 in above-mentioned memory module 1 is to patrolling
The human task for collecting processing module is pre-processed, and obtains n candidate result, wherein the number of candidate result can be according to difference
Demand is voluntarily configured or artificial intelligence system is adaptive；
In present pre-ferred embodiments, in above-mentioned artificial intelligence task processing system, controller 12 carries out pretreatment and appoints
When business or processing related data, the intelligent algorithm used includes the intelligent algorithm compressed or unpressed artificial intelligence
Algorithm；Wherein, the intelligent algorithm of the compression include low accuracy data type expression parameter intelligent algorithm and
Deleted the intelligent algorithm of neuron；The unpressed intelligent algorithm include high-precision data expression parameter and
The intelligent algorithm of neuron was not deleted.
In present pre-ferred embodiments, in above-mentioned artificial intelligence task processing system, the data that controller 12 uses, packet
The data and unpressed data of compression are included, wherein the data compressed include data for the picture cut out and using compress technique
Processed data
Further, when above-mentioned logic processing module 2 is worked, it is only necessary to be calculated from the controller 12
Most correct result is filtered out in n candidate result or is further processed.
In present pre-ferred embodiments, for the specific value of n, it can be set as required depending on the situation
Fixed, after setting respective value, pretreated result can generate the candidate result of corresponding number as needed；Wherein preset n value
Bigger, the result of acquisition is more accurate.
In present pre-ferred embodiments, 2 free time of logic processing module, such as night, in storage element are selected
Data such as picture carry out recognition of face, the working condition of the artificial intelligence task processing system is as shown in Figure 2:
Controller in memory module receives the compressed data of storage unit；
Further, the controller in memory module such as uses using the neural network model of compression and utilizes 16 floating-points
The supplemental characteristic that the variable of type indicates, tentatively to be identified to compressed picture.
Further, above-mentioned identification process are as follows: controller can be according to the n of preset or default result number
Value, the compressed picture of input is tentatively identified, n candidate result is obtained；
Further, when logic processing module state of resuming work is, logic processing module will do it calculating, and from above-mentioned
It is picked out in n candidate result most accurately as a result, realizing accurate recognition of face.
In conclusion specific steps such as Fig. 3 in present pre-ferred embodiments, when artificial intelligence task processing system works
It is shown:
Step S1: the data that input is handled start intelligent algorithm；
Step S2: whether decision logic processing module is in running order；
If being judged as YES, directly progress step S4；
If being judged as NO, next step S3 is carried out；
Step S3: the controller in memory module pre-processes task, and according to preset processing result number n,
Export n candidate result；
Step S4: logic processing module carries out processing to artificial intelligence task or according to the n time calculated from step S3
It selects and selects optimal result in result.
In present pre-ferred embodiments, in above-mentioned artificial intelligence task processing system, when logic processing module 2 is idle,
It is handled when working without waiting for logic processing module 2, but is pre-processed by controller 12 in memory module 1 again, it is real
Now efficient, energy-efficient artificial intelligence task processing.
The foregoing is merely preferred embodiments of the present invention, are not intended to limit embodiments of the present invention and protection model
It encloses, to those skilled in the art, should can appreciate that all with made by description of the invention and diagramatic content
Equivalent replacement and obviously change obtained scheme, should all be included within the scope of the present invention.
Claims (9)
1. a kind of artificial intelligence task processing system, which is characterized in that including memory module and logic processing module, the logic
Processing module is connected with the storage storing module；
The memory module includes storage unit and the controller for connecting the storage unit, is saved in the storage unit related
It is coupled to the initial data of artificial intelligence task；
The neural network model for being handled the initial data is preset in the controller, when the logical process
When module is not in working condition, the controller is handled the initial data using the neural network model, and
Export multiple processing results；
When the logic processing module is in running order, the logic processing module exports multiple according to the controller
The processing result is further processed, to obtain the final process result and the output that are associated with the artificial intelligence task.
2. artificial intelligence task processing system as described in claim 1, which is characterized in that the logic processing module is to multiple
The mode that the processing result is further processed are as follows: screening obtains accuracy highest one in multiple processing results
It is a and as the final process result export.
3. artificial intelligence task processing system as described in claim 1, which is characterized in that preset described in the controller
Neural network model is the neural network model through overcompression.
4. artificial intelligence task processing system as described in claim 1, which is characterized in that the institute saved in the storage unit
Stating initial data is the initial data through overcompression.
5. artificial intelligence task processing system as described in claim 1, which is characterized in that the memory module is non-volatile
Memory.
6. artificial intelligence task processing system as described in claim 1, which is characterized in that the controller be microcontroller or
Microprocessor.
7. artificial intelligence task processing system as described in claim 1, which is characterized in that in Yu Suoshu controller, set in advance
Surely the quantity of the processing result exported.
8. a kind of artificial intelligence task processing method, which is characterized in that applied to as described in any one of claim 1-7
Artificial intelligence task processing system, further includes:
When step S1, Yu Suoshu logic processing module is not at working condition, the controller in the storage unit to saving
The initial data for being associated with the artificial intelligence task handled, and export multiple processing results；
Step S2, when the logic processing module is in running order, the logic processing module is defeated according to the controller
Multiple processing results out are further processed, to obtain being associated with the final process result of the artificial intelligence task
And it exports.
9. artificial intelligence task processing method as claimed in claim 8, which is characterized in that in the step S2, the logic
The mode that multiple processing results are further processed in processing module are as follows: screen and obtain in multiple processing results
Accuracy it is highest one and as the final process result export.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811628648.8A CN109656721A (en) | 2018-12-28 | 2018-12-28 | A kind of efficient intelligence system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811628648.8A CN109656721A (en) | 2018-12-28 | 2018-12-28 | A kind of efficient intelligence system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109656721A true CN109656721A (en) | 2019-04-19 |
Family
ID=66116912
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811628648.8A Pending CN109656721A (en) | 2018-12-28 | 2018-12-28 | A kind of efficient intelligence system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109656721A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117312388A (en) * | 2023-10-08 | 2023-12-29 | 江苏泰赋星信息技术有限公司 | Artificial intelligence model control system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104346214A (en) * | 2013-07-30 | 2015-02-11 | 中国银联股份有限公司 | Device and method for managing asynchronous tasks in distributed environments |
| US20160045162A1 (en) * | 2008-10-07 | 2016-02-18 | Mc10, Inc. | Systems, methods, and devices having stretchable integrated circuitry for sensing and delivering therapy |
| CN105828041A (en) * | 2016-04-11 | 2016-08-03 | 上海大学 | Video acquisition system supporting parallel preprocessing |
| CN109086867A (en) * | 2018-07-02 | 2018-12-25 | 武汉魅瞳科技有限公司 | A kind of convolutional neural networks acceleration system based on FPGA |
-
2018
- 2018-12-28 CN CN201811628648.8A patent/CN109656721A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160045162A1 (en) * | 2008-10-07 | 2016-02-18 | Mc10, Inc. | Systems, methods, and devices having stretchable integrated circuitry for sensing and delivering therapy |
| CN104346214A (en) * | 2013-07-30 | 2015-02-11 | 中国银联股份有限公司 | Device and method for managing asynchronous tasks in distributed environments |
| CN105828041A (en) * | 2016-04-11 | 2016-08-03 | 上海大学 | Video acquisition system supporting parallel preprocessing |
| CN109086867A (en) * | 2018-07-02 | 2018-12-25 | 武汉魅瞳科技有限公司 | A kind of convolutional neural networks acceleration system based on FPGA |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117312388A (en) * | 2023-10-08 | 2023-12-29 | 江苏泰赋星信息技术有限公司 | Artificial intelligence model control system |
| CN117312388B (en) * | 2023-10-08 | 2024-03-19 | 江苏泰赋星信息技术有限公司 | Artificial intelligence model control system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105157087B (en) | The control method and device of source pump | |
| CN105278353B (en) | Method, system and the data processing equipment of intelligent acquisition data | |
| CN104915247B (en) | A kind of real time data calculation method and system | |
| CN108133391A (en) | Method for Sales Forecast method and server | |
| EP3514647A3 (en) | System and method for reconfiguring a system-on-module for an unmanned vehicle | |
| CN111008640A (en) | Image recognition model training and image recognition method, device, terminal and medium | |
| CN110427419A (en) | A kind of flying quality anomalous identification model building method based on data mining | |
| CN104216783A (en) | Method for automatically managing and controlling virtual GPU (Graphics Processing Unit) resource in cloud gaming | |
| CN109656721A (en) | A kind of efficient intelligence system | |
| CN111651220B (en) | Spark parameter automatic optimization method and system based on deep reinforcement learning | |
| CN111831355B (en) | Weight precision configuration method, device, equipment and storage medium | |
| CN110232436A (en) | Pruning method, device and the storage medium of convolutional neural networks | |
| CN113010312B (en) | Super-parameter tuning method, device and storage medium | |
| CN107861458A (en) | It is a kind of can autonomous configuration hardware resource PLC fast construction methods | |
| CN109359724A (en) | A kind of method and device of compression storage convolution neural network model | |
| CN108629627A (en) | A kind of automatic dispensing control method of moving advertising | |
| CN112200391A (en) | Power distribution network edge side load prediction method based on k-nearest neighbor mutual information characteristic simplification | |
| CN117090758A (en) | Energy-saving control method and system for air compressor | |
| CN117130888A (en) | Automatic server tuning method and system based on historical data and machine learning | |
| CN107743331A (en) | Light source control method, device and system | |
| CN110276452A (en) | Pruning method, device, equipment and the artificial intelligence chip of neural network model | |
| CN115794382A (en) | Data processing method, device and nonvolatile storage medium | |
| CN116415714A (en) | Wind power prediction method and device, electronic equipment and readable storage medium | |
| CN105843891A (en) | Incremental online characteristic extraction and analysis method and system | |
| US11846278B2 (en) | Systems and methods for centralized optimization of reservoir production |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190419 |