CN109284829A - Recognition with Recurrent Neural Network based on evaluation network - Google Patents
Recognition with Recurrent Neural Network based on evaluation network Download PDFInfo
- Publication number
- CN109284829A CN109284829A CN201811115285.8A CN201811115285A CN109284829A CN 109284829 A CN109284829 A CN 109284829A CN 201811115285 A CN201811115285 A CN 201811115285A CN 109284829 A CN109284829 A CN 109284829A
- Authority
- CN
- China
- Prior art keywords
- evaluation
- network
- recognition
- recurrent neural
- neural network
- 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
- 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)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the Recognition with Recurrent Neural Network based on evaluation network comprising the steps of: A, does numeralization processing to the training sample in RNN；B, the network structure of RNN is determined；C, the input neuron number of evaluation network is determined；D, the output neuron number of evaluation network is determined；E, the number of evaluation network is determined；F, train containing evaluation index to training sample；G, it is predicted using the RNN based on evaluation network after training.Recognition with Recurrent Neural Network and evaluation network are combined by the present invention, it is put into evaluation network (based on artificial neural network) using the result that Recognition with Recurrent Neural Network obtains, and it is evaluated, the result of evaluation is fed back into Recognition with Recurrent Neural Network, so that Recognition with Recurrent Neural Network is after the certain sample of training, can obtain meeting some evaluation index (such as practicability, reliability or aesthetics etc.) as a result, improving the availability of Recognition with Recurrent Neural Network.
Description
Technical field
The present invention relates to field of artificial intelligence, specifically the Recognition with Recurrent Neural Network based on evaluation network.
Background technique
Recognition with Recurrent Neural Network is suggested, maximum feature to solve the problems, such as study things timing, can be with
Learn the timing of things using Recognition with Recurrent Neural Network, and is carried out intentionally according to its content (timing of content) learnt
The combination of justice, therefore it is used for the creation of article, the creation of ci and qu etc. more, is artificial intelligence field machine learning
One very important algorithm.
However utilize it is trained after Recognition with Recurrent Neural Network in reasoning, can only be according in the previously sample that had learnt
Timing make inferences, although obtained result can have preferable correlation with the timing of things in previous sample,
It is its result for obtaining is only in some level as a result, this will lead to the result of its generation and contemplates that certain gap.
After such as being learnt according to a certain amount of sample (such as poem), the result obtained when making inferences often has with previous sample
Correlation, but as a result, no rhymed, if there is artistic conception etc., the result that can not be embodied, and obtained well
Compare difficult and previous sample and form closed loop, therefore allows for its application scenarios and be subject to certain restrictions.
Summary of the invention
The purpose of the present invention is to provide the Recognition with Recurrent Neural Network based on evaluation network, to solve to mention in above-mentioned background technique
Out the problem of.
To achieve the above object, the invention provides the following technical scheme:
Recognition with Recurrent Neural Network based on evaluation network comprising the steps of:
A, numeralization processing is done to the training sample in RNN；
B, the network structure of RNN is determined；
C, the input neuron number of evaluation network is determined；
D, the output neuron number of evaluation network is determined；
E, the number of evaluation network is determined；
F, train containing evaluation index to training sample；
G, it is predicted using the RNN based on evaluation network after training.
As further technical solution of the present invention: the step A is specifically: neural to circulation is put into using various ways
Training sample in network carries out numeralization processing, including but not limited to the side such as sample vector, word2vec, TF-IDF
Formula carries out numeralization processing to the content of output.
As further technical solution of the present invention: the step B is specifically: according to the processing side to quantize in step A
Formula determines the neuron number output and input in a neural network in circulation RNN, and according to sample to be trained
This, to determine the cycle-index of Recognition with Recurrent Neural Network RNN.
As further technical solution of the present invention: the step C is specifically: according to the circulation of the RNN determined in step B
Number inputs the circulation time of neuron number and Recognition with Recurrent Neural Network to determine the input neuron number in evaluation network
Number is identical, i.e. the output result of realization Recognition with Recurrent Neural Network is mapped in evaluation network.
As further technical solution of the present invention: in the step D, evaluating the output neuron number of network, foundation
Different appraisement systems and different scoring tactics provide.
As further technical solution of the present invention: in the step E, evaluation index is divided into different evaluations depending on the application
Network including but not limited to evaluates commenting for the evaluation network of practicability, the evaluation network for evaluating appearance and evaluation reliability etc.
Valence network, evaluation index are also different and different with different evaluation informations.
As further technical solution of the present invention: the step F is specifically: it is different according to the index of evaluation, to being put into
Sample in RNN based on evaluation network provides corresponding scoring, and will be put into training sample and scoring based on evaluation net
It is trained in the RNN of network.
As further technical solution of the present invention: the step G is specifically: based on evaluation network RNN complete to
After the training of training sample, user is predicted using the RNN based on evaluation network, and the content predicted will be needed to be put into base
After the RNN of evaluation network, can be provided according to previous trained network the timing based on the content as a result, and to
Scoring of the result in different evaluation index out.
Compared with prior art, the beneficial effects of the present invention are: the present invention carries out Recognition with Recurrent Neural Network and evaluation network
In conjunction with, it is put into evaluation network (based on artificial neural network) using the result that Recognition with Recurrent Neural Network obtains, and evaluated,
The result of evaluation is fed back into Recognition with Recurrent Neural Network, so that Recognition with Recurrent Neural Network can obtain after the certain sample of training
To meet some evaluation index (such as practicability, reliability or aesthetics etc.) as a result, improving Recognition with Recurrent Neural Network
Availability.
Detailed description of the invention
Fig. 1 is circulation (recurrence) neural network diagram based on evaluation network；
Fig. 2 is the Recognition with Recurrent Neural Network design method figure based on evaluation network.
Specific embodiment
The technical scheme in the embodiments of the invention will be clearly and completely described below, it is clear that described implementation
Example is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this field is common
Technical staff's every other embodiment obtained without making creative work belongs to the model that the present invention protects
It encloses.
Recognition with Recurrent Neural Network referring to FIG. 1-2, based on evaluation network comprising the steps of:
A, numeralization processing is done to the training sample in RNN.Training of the various ways to being put into Recognition with Recurrent Neural Network can be used
Sample carries out numeralization processing, including but not limited to sample vector, word2vec, TF-IDF etc. mode to output
Content carries out numeralization processing；
B, the network structure of RNN is determined.According to the processing mode to quantize in step 1, a nerve net in circulation RNN is determined
The neuron number output and input in network.And according to sample to be trained, to determine Recognition with Recurrent Neural Network (RNN)
Cycle-index；
C, the input neuron number of evaluation network is determined.According to the cycle-index of the RNN determined in step 2, to determine evaluation
Input neuron number in network, input neuron number is identical as the cycle-index of Recognition with Recurrent Neural Network, that is, realizes and follow
The output result of ring neural network is mapped in evaluation network (as shown in Figure 1)；
D, the output neuron number of evaluation network is determined.The output neuron number for evaluating network, according to different evaluation bodies
System and different scoring tactics provide, (for example scoring tactics are 5 points of systems, then the number for evaluating the output neuron of network is 5
It is a；If it is hundred-mark system, the output neuron number for evaluating network is 100)；
E, the number of evaluation network is determined.Based on evaluation network RNN in, evaluation network can have it is multiple, with realize to not
With the evaluation of index.Evaluation index can be divided into different evaluation networks depending on the application, such as evaluate the evaluation network of practicability, comment
The evaluation network of valence appearance and the evaluation network etc. for evaluating reliability, can be divided into different evaluation networks, evaluation index
Also different and different (as shown in Figure 2) with different evaluation informations；
F, train containing evaluation index to training sample.It is different according to the index of evaluation, to being put into the RNN based on evaluation network
In sample provide corresponding scoring, and be trained being put into the RNN based on evaluation network to training sample and scoring；
G, it is predicted using the RNN based on evaluation network after training.RNN based on evaluation network is completed to training sample
Training after, user can use the RNN based on evaluation network and predict, the content predicted will be needed to be put into based on commenting
After the RNN of valence network, can be provided according to previous trained network the timing based on the content as a result, and providing this
As a result the scoring in different evaluation index (such as practicability, appearance, reliability etc.), so that user be helped to fully understand
The result of RNN.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (8)
1. the Recognition with Recurrent Neural Network based on evaluation network, which is characterized in that comprise the steps of:
A, numeralization processing is done to the training sample in RNN；
B, the network structure of RNN is determined；
C, the input neuron number of evaluation network is determined；
D, the output neuron number of evaluation network is determined；
E, the number of evaluation network is determined；
F, train containing evaluation index to training sample；
G, it is predicted using the RNN based on evaluation network after training.
2. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that the step A is specific
It is: numeralization processing is carried out to the training sample being put into Recognition with Recurrent Neural Network using various ways, including sample vector
Change, word2vec, TF-IDF mode carry out numeralization processing to the content of output.
3. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that the step B is specific
It is: according to the processing mode to quantize in step A, determines outputting and inputting in a neural network in circulation RNN
Neuron number, and according to sample to be trained, to determine the cycle-index of Recognition with Recurrent Neural Network RNN.
4. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that the step C is specific
It is: according to the cycle-index of the RNN determined in step B, to determine the input neuron number in evaluation network, input nerve
First number is identical as the cycle-index of Recognition with Recurrent Neural Network, i.e. the output result of realization Recognition with Recurrent Neural Network is mapped to evaluation network
In.
5. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that in the step D, comment
The output neuron number of valence network is provided according to different appraisement systems and different scoring tactics.
6. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that in the step E, comment
Valence index is divided into different evaluation networks depending on the application, the evaluation network of evaluation network, evaluation appearance including evaluation practicability
And the evaluation network of evaluation reliability, evaluation index are also different and different with different evaluation informations.
7. the Recognition with Recurrent Neural Network according to claim 1 based on evaluation network, which is characterized in that the step F is specific
It is: it is different according to the index of evaluation, corresponding scoring is provided to the sample being put into the RNN based on evaluation network, and will be wait instruct
Practice sample and scoring is put into the RNN based on evaluation network and is trained.
8. -7 any Recognition with Recurrent Neural Network based on evaluation network according to claim 1, which is characterized in that the step
G is specifically: the RNN based on evaluation network is completed after the training of training sample, and user utilizes the RNN based on evaluation network
It is predicted, it, can be according to previous trained network after needing the content predicted to be put into the RNN based on evaluation network
Provide the timing based on the content as a result, and providing scoring of the result in different evaluation index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811115285.8A CN109284829A (en) | 2018-09-25 | 2018-09-25 | Recognition with Recurrent Neural Network based on evaluation network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811115285.8A CN109284829A (en) | 2018-09-25 | 2018-09-25 | Recognition with Recurrent Neural Network based on evaluation network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109284829A true CN109284829A (en) | 2019-01-29 |
Family
ID=65181283
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811115285.8A Pending CN109284829A (en) | 2018-09-25 | 2018-09-25 | Recognition with Recurrent Neural Network based on evaluation network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109284829A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738313A (en) * | 2019-10-15 | 2020-01-31 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for evaluating quantization operation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170178346A1 (en) * | 2015-12-16 | 2017-06-22 | High School Cube, Llc | Neural network architecture for analyzing video data |
| CN106951783A (en) * | 2017-03-31 | 2017-07-14 | 国家电网公司 | A kind of Method for Masquerade Intrusion Detection and device based on deep neural network |
| CN107122790A (en) * | 2017-03-15 | 2017-09-01 | 华北电力大学 | Non-intrusion type load recognizer based on hybrid neural networks and integrated study |
| CN108197702A (en) * | 2018-02-09 | 2018-06-22 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of method of the product design based on evaluation network and Recognition with Recurrent Neural Network |
| CN108363697A (en) * | 2018-03-08 | 2018-08-03 | 腾讯科技（深圳）有限公司 | Text message generation method, device, storage medium and equipment |
-
2018
- 2018-09-25 CN CN201811115285.8A patent/CN109284829A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170178346A1 (en) * | 2015-12-16 | 2017-06-22 | High School Cube, Llc | Neural network architecture for analyzing video data |
| CN107122790A (en) * | 2017-03-15 | 2017-09-01 | 华北电力大学 | Non-intrusion type load recognizer based on hybrid neural networks and integrated study |
| CN106951783A (en) * | 2017-03-31 | 2017-07-14 | 国家电网公司 | A kind of Method for Masquerade Intrusion Detection and device based on deep neural network |
| CN108197702A (en) * | 2018-02-09 | 2018-06-22 | 艾凯克斯（嘉兴）信息科技有限公司 | A kind of method of the product design based on evaluation network and Recognition with Recurrent Neural Network |
| CN108363697A (en) * | 2018-03-08 | 2018-08-03 | 腾讯科技（深圳）有限公司 | Text message generation method, device, storage medium and equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738313A (en) * | 2019-10-15 | 2020-01-31 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for evaluating quantization operation |
| CN110738313B (en) * | 2019-10-15 | 2022-05-31 | 阿波罗智能技术(北京)有限公司 | Method, apparatus, device and medium for evaluating quantization operation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107437077A (en) | A kind of method that rotation face based on generation confrontation network represents study | |
| Hidayah et al. | Student classification for academic performance prediction using neuro fuzzy in a conventional classroom | |
| Suresh et al. | Using transformers to provide teachers with personalized feedback on their classroom discourse: The TalkMoves application | |
| Pérez et al. | Sustainable development seen from environmental training in university linkage | |
| CN105701540A (en) | Self-generated neural network construction method | |
| CN109388698A (en) | A kind of guiding automatic chatting method based on deeply study | |
| CN108364066B (en) | Artificial neural network chip and its application method based on N-GRAM and WFST model | |
| Blomkamp | A critical history of cultural indicators | |
| Singh | Colorization of old gray scale images and videos using deep learning | |
| Wolhuter | Research on HE in South Africa: stocktaking and assessment from international comparative perspectives | |
| Feng et al. | Research on the multimodal digital teaching quality data evaluation model based on fuzzy BP neural network | |
| Wildenberg et al. | Linking thoughts to flows-Fuzzy cognitive mapping as tool for integrated landscape modelling | |
| CN109284829A (en) | Recognition with Recurrent Neural Network based on evaluation network | |
| Praharaj | Co-located collaboration analytics | |
| US11727338B2 (en) | Controlling submission of content | |
| Das et al. | Application of fuzzy logic in the ranking of academic institutions | |
| Yu et al. | Temporal sentiment analysis of learners: Public versus private social media communication channels in a women-in-tech conversion course | |
| Chetouani et al. | Human-centered artificial intelligence: Advanced lectures | |
| Kuptsov et al. | Modeling of pedagogical processes | |
| Tuyen et al. | Forecasting nonverbal social signals during dyadic interactions with generative adversarial neural networks | |
| Piedra Calderón et al. | Global collective intelligence in technological societies: as a result of collaborative knowledge in combination with artificial intelligence | |
| Massel et al. | Contingency management and semantic modeling in energy sector | |
| Yanchao et al. | Research on undergraduate academic prediction model based on deep learning | |
| Chernyakov et al. | Improving the financial stability of economic entities in the agricultural sector | |
| Gupta et al. | Per-C based student examination strategy evaluation in mobile evaluation system conducted through a smartphone |
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: 20190129 |