CN109389207A - A kind of adaptive neural network learning method and nerve network system - Google Patents
A kind of adaptive neural network learning method and nerve network system Download PDFInfo
- Publication number
- CN109389207A CN109389207A CN201811173901.5A CN201811173901A CN109389207A CN 109389207 A CN109389207 A CN 109389207A CN 201811173901 A CN201811173901 A CN 201811173901A CN 109389207 A CN109389207 A CN 109389207A
- Authority
- CN
- China
- Prior art keywords
- output
- layer
- neural network
- knowledge
- similarity
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 22
- 210000005036 nerve Anatomy 0.000 title abstract 3
- 230000006870 function Effects 0.000 claims abstract description 9
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of adaptive neural network learning method and the nerve network systems, the nerve network system includes: input layer, hidden layer, output layer and knowledge base, this method includes: (1) creating and initialize knowledge base, the weighed value adjusting to hidden layer and output layer simultaneously makes output result similarity meet the condition of convergence；(2) online adaptive learns, and obtains output result using learning algorithm using this group of connection weight as initial value for a certain input data；(3) judge whether export result and the similarity of corresponding desired output meets the requirements: when demanded, then exporting result；Otherwise, step (2) are repeated, goes through all knowledge in knowledge base, does not find yet, then regard the data as the sample of a new knowledge；(4) connection weight is adjusted according to step (1), is added in knowledge base.Method of the invention is capable of handling and differentiates inside and outside data, new legacy data, realizes online adaptive study and identification function.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to a self-adaptive neural network learning method and a neural network system.
Background
Within the discipline of artificial intelligence, the research efforts of neural networks have been successfully transplanted into a considerable number of fields, such as decision support, face recognition, knowledge base systems, expert systems, and emotional robots. In representing traditional research, most models generally work in already explained areas: that is, for an interpreted context, the system designer typically gives some implicit a priori conventions under which it is difficult to transform the context, objective, or representation as the problem solving process progresses.
At present, the main means of the image thinking simulation is to simulate the connection mechanism mainly based on the 'intelligent' -artificial neural network connection related to the copied image thinking. In terms of a calculation processing method, a new approach is opened up by a connection mechanism method, namely a method of parallel processing and distributed expression is adopted. Specifically, the method uses a network of a plurality of nodes, each of which can be connected to each other, to represent information. The semantic network used for expressing knowledge in the past is that one node corresponds to one concept, and the artificial neural network corresponds to one concept by a distribution mode of the nodes and the size of a weighting quantity, so that even if the information attribute on each node is distorted and distorted, the attribute of the concept expressed by the network is not changed greatly. In addition, information on some common elements can also be used to express similar concepts, but simulating visual thinking in various methods including the neural network connection method described above is also not completely successful, as well as symbolic representations of logical thinking.
The neural network system can identify the input, and the input related knowledge and grammar rules are expressed by the neural network structure and the neuron connection weight, so that the neural network system has even a certain degree of fault tolerance. Neural network systems are data driven and it is unclear whether these data sources are internal or external, new or old. After the neural network is trained, if the input data into the system is data not contained in the training set, namely a new data sample, the neural network cannot judge whether the input is new information or not relative to the knowledge of the input data, and cannot actively learn the new data, but the knowledge obtained in the training process is used for carrying out error judgment on the new data sample. Or, in the online training process, the neural network does not distinguish the knowledge states which are already existed by the neural network, and the knowledge states which are externally input data are always processed without difference.
Disclosure of Invention
The invention aims to provide a self-adaptive neural network learning method and a neural network system, which solve the problem that the existing neural network system cannot distinguish new data from old data and learn, can process and distinguish internal data and external data and new data from old data, and realize the functions of on-line self-adaptive learning and identification.
In order to achieve the above object, the present invention provides an adaptive neural network learning method, wherein a neural network system comprises: input layer, hidden layer, output layer and knowledge base, where K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkbWhere b is 1,2 … … m, Y is the desired output, and V and W are the connection weight of the network hidden layer and the connection weight of the output layer, respectively.
The method comprises the following steps:
(1) establishing and initializing a knowledge base: acquiring knowledge of a training set, wherein each knowledge corresponds to a group of connection weights V of a hidden layer and a group of connection weights W of an output layer, adjusting the weights of the hidden layer and the output layer in the training process, enabling the similarity E of output results to meet a convergence condition, and determining the final output connection weight;
(2) on-line adaptive learning, searching any knowledge in the knowledge base to obtain a connection weight V of a corresponding hidden layer and a connection weight W of an output layer, taking the group of connection weights as an initial value for certain input data, adjusting the connection weights by using the adjustment of the weights in the step 1, and continuously calculating the specified learning times N to obtain an output result of actual operation;
(3) judging whether the similarity of the output result and the corresponding expected output meets the requirement: when the similarity between the output result and the corresponding expected output meets the requirement, outputting the result; when the similarity between the output result and the corresponding expected output does not meet the requirement, searching and selecting a new knowledge from the knowledge base in sequence, acquiring a new set of hidden layer, output layer and expected output, repeating the step (2), and if all the knowledge in the knowledge base is gone through and the corresponding knowledge is not found yet, regarding the data as a sample of a new knowledge;
(4) and (3) adjusting the connection weight according to the adjustment of the weight in the step (1) until the similarity E of the output result meets the convergence condition, combining the connection weight of the corresponding output layer, the connection weight of the hidden layer and the expected output into a new knowledge, and adding the new knowledge into a knowledge base K.
Preferably, the connection weight of the network hidden layer and the connection weight of the output layer are both matrixes; wherein,
preferably, in step (1), the algorithm for adjusting the weight includes:
δj＝(yj-oj)f′(netj) (3)。
in formulae (1) to (3), wijAnd vijThe connection weights of the output layer and the hidden layer in the matrix (i, j), α are the scaling factors, deltajIs the learning rate of j columns in the matrix, f' (net)j) As the derivative of the neuron excitation function, yjAnd ojDesired output and actual output, o, respectively, of j columns in the matrixiIs the actual output of the i rows in the matrix.
Preferably, in step (1), the equation of the convergence condition of the output result similarity E is:
E＝∑Ep(5)。
in formulae (4) and (5), EpRepresenting the resultant similarity, y, of the pp-th output neuronPjAnd oPjRespectively, the expected output and the actual output of (p, j) in the matrix, and E represents the result similarity and is used for judging whether the network meets the convergence requirement.
Preferably, in the step (1), the similarity requirement is satisfied when the result similarity E is lower than 20% of the expected output result; in step (4), the similarity requirement is satisfied when the result similarity E is lower than 5% of the expected output result.
The present invention also provides an adaptive neural network system, comprising: input layer, hidden layer, output layer and knowledge base, where K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkb(V | W | Y), b ═ 1,2 … … m, Y is the desired output, V and W are the connection weights of the network hidden layer and the output layer, respectively; the neural network system is a forward type neural network system, the input layer, the hidden layer and the output layer are sequentially connected for transmission, and the knowledge base is connected with the hidden layer.
Preferably, the connection weight of the network hidden layer and the connection weight of the output layer are both matrixes; wherein,
the self-adaptive neural network learning method and the neural network system solve the problem that the existing neural network system cannot distinguish new data from old data and learn, and have the following advantages that:
the self-adaptive neural network learning method and the neural network system store the neural network connection weight in the knowledge base in the form of knowledge, thereby realizing the processing of searching, adding and the like of the knowledge, leading the neural network system to have the functions of processing and distinguishing internal and external data and new and old data, and further realizing the functions of on-line self-adaptive learning and identification. Moreover, the method of the present invention is flexible in transforming context, goals or representations for the context of the interpretation as the problem-solving process progresses.
Drawings
FIG. 1 is a schematic diagram of a neural network system of the present invention.
FIG. 2 is a flow chart of the adaptive neural network learning method of the present invention.
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
A method for adaptive neural network learning, as shown in fig. 1, is a schematic diagram of a neural network system of the present invention, the neural network system comprising: input layer LA＝(a1……ah……an) Hidden layer LB＝(b1……bi……bp) An output layer LC＝(c1……cj……cq) And knowledge base K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkbWhere (V | W | Y), b ═ 1,2 … … m (natural numbers), Y is the desired output, and V and W are the connection weights of the network hidden layer and the output layer, respectively, where,
as shown in fig. 2, a flow chart of the adaptive neural network learning method of the present invention is shown, the method includes:
(1) establishing and initializing a knowledge base: for each data set, namely training set, sent to the input layer, a knowledge can be obtained, namely the connection weight V of a group of corresponding hidden layers and the connection weight W of the output layer, and the weight adjustment algorithm of the hidden layers and the output layer in the learning process is as follows:
δj＝(yj-oj)f′(netj) (3)；
in formulae (1) to (3), wijAnd vijThe connection weights of the output layer and the hidden layer in the matrix (i, j), α are the scaling factors, deltajIs the learning rate of j columns in the matrix, f' (net)j) As the derivative of the neuron excitation function, yjAnd ojDesired output and actual output, o, respectively, of j columns in the matrixiIs the actual output of the i rows in the matrix.
Determining the final output connection weight value according to whether the output result similarity E meets the convergence condition, namely determining one knowledge in a knowledge base:
E＝∑Ep(5)；
in formulae (4) and (5), EpIs shown asResult similarity, y, of pp output neuronsPjAnd oPjRespectively, the expected output and the actual output of (p, j) in the matrix, and E represents the result similarity and is used for judging whether the network meets the convergence requirement.
(2) On-line adaptive learning, setting learning times N, setting output result similarity, namely an error E value, searching any knowledge in a knowledge base, separating a connection weight V of a corresponding hidden layer and a connection weight W of an output layer, applying the group of connection weights as an initial value to certain input data, adjusting the connection weights by using the adjustment of the weights in the step (1), and continuously calculating the specified learning times N to obtain an output result of actual operation.
(3) Judging whether the similarity of the output result and the corresponding expected output meets the requirement: if the similarity between the output result and the corresponding expected output meets the requirement, outputting the result, and if the result similarity E is lower than 20% of the expected output result, meeting the requirement of the similarity; if the similarity between the output result and the corresponding expected output does not meet the requirement, searching and selecting a new knowledge from the knowledge base in sequence, namely selecting a new set of hidden layer, output layer and expected output, repeating the step (2), if all the knowledge in the knowledge base is passed, the corresponding knowledge is not found, and even if the similarity between the output result and the corresponding expected output meets the requirement, the data is regarded as a sample of a new knowledge;
(4) and (3) adjusting the connection weight according to the formulas (1) to (3) in the step (1) until the similarity E of the output result meets the convergence condition, combining the connection weight of the corresponding output layer, the connection weight of the hidden layer and the expected output into a new knowledge in order to meet the similarity requirement when the error is lower than 5% of the expected output result, and adding the new knowledge into a knowledge base K.
An adaptive neural network system, as shown in fig. 1, is a schematic diagram of the neural network system of the present invention, the neural network system is a forward type neural network system, and includes: input layer LA＝(a1……ah……an) Hidden layerLB＝(b1……bi……bp) An output layer LC＝(c1……cj……bq) And knowledge base K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkb(V | W | Y), b ═ 1,2 … … m, Y is the desired output, V and W are the connection weights of the hidden layer and the output layer of the network, respectively, the input layer, the hidden layer and the output layer are connected in sequence for transmission, the knowledge base is connected to the hidden layer, wherein,
in summary, the adaptive neural network learning method and the neural network system of the present invention store the neural network connection weight in the knowledge base in the form of knowledge, thereby implementing processes such as searching and adding knowledge, and enabling the neural network system to have functions of processing and distinguishing internal and external data and new and old data, thereby implementing functions of online adaptive learning and identification.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.
Claims (7)
1. An adaptive neural network learning method, wherein a neural network system comprises: input layer, hidden layer, output layer and knowledge base, where K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkb(V | W | Y), b ═ 1,2 … … m, Y is the desired output, V and W are the connection weights of the network hidden layer and the output layer, respectively;
the method comprises the following steps:
(1) establishing and initializing a knowledge base: acquiring knowledge of a training set, wherein each knowledge corresponds to a group of connection weights V of a hidden layer and a group of connection weights W of an output layer, adjusting the weights of the hidden layer and the output layer in the training process, enabling the similarity E of output results to meet a convergence condition, and determining the final output connection weight;
(2) on-line adaptive learning, searching any knowledge in the knowledge base to obtain a connection weight V of a corresponding hidden layer and a connection weight W of an output layer, taking the group of connection weights as an initial value for certain input data, adjusting the connection weights by using the adjustment of the weights in the step 1, and continuously calculating the specified learning times N to obtain an output result of actual operation;
(3) judging whether the similarity of the output result and the corresponding expected output meets the requirement: when the similarity between the output result and the corresponding expected output meets the requirement, outputting the result; when the similarity between the output result and the corresponding expected output does not meet the requirement, searching and selecting a new knowledge from the knowledge base in sequence, acquiring a new set of hidden layer, output layer and expected output, repeating the step (2), and if all the knowledge in the knowledge base is gone through and the corresponding knowledge is not found yet, regarding the data as a sample of a new knowledge;
(4) and (3) adjusting the connection weight according to the adjustment of the weight in the step (1) until the similarity E of the output result meets the convergence condition, combining the connection weight of the corresponding output layer, the connection weight of the hidden layer and the expected output into a new knowledge, and adding the new knowledge into a knowledge base K.
2. The adaptive neural network learning method of claim 1, wherein the connection weights of the hidden layer and the output layer are matrices; wherein,
3. the adaptive neural network learning method of claim 2, wherein in step (1), the weight value adjustment algorithm comprises:
δj＝(yj-oj)f′(netj) (3)；
in formulae (1) to (3), wijAnd vijThe connection weights of the output layer and the hidden layer in the matrix (i, j), α are the scaling factors, deltajIs the learning rate of j columns in the matrix, f' (net)j) As the derivative of the neuron excitation function, yjAnd ojDesired output and actual output, o, respectively, of j columns in the matrixiIs the actual output of the i rows in the matrix.
4. The adaptive neural network learning method according to any one of claims 1 to 3, wherein in step (1), the equation of the convergence condition of the output result similarity E is:
E＝∑Ep(5)；
in formulae (4) and (5), EpRepresenting the resulting similarity, y, of the p-th output neuronPjAnd oPjRespectively, the expected output and the actual output of (p, j) in the matrix, and E represents the result similarity and is used for judging whether the network meets the convergence requirement.
5. The adaptive neural network learning method of claim 1, wherein in step (1), the similarity requirement is satisfied when the result similarity E is lower than 20% of the expected output result; in step (4), the similarity requirement is satisfied when the result similarity E is lower than 5% of the expected output result.
6. An adaptive neural network system, comprising: input layer, hidden layer, output layer and knowledge base, where K ═ S1,S2……Sm)，SbTo connect the weight, S, to the neural networkb(V | W | Y), b ═ 1,2 … … m, Y is the desired output, V and W are the connection weights of the network hidden layer and the output layer, respectively; the neural network system is a forward type neural network system, the input layer, the hidden layer and the output layer are sequentially connected for transmission, and the knowledge base is connected with the hidden layer.
7. The adaptive neural network system of claim 6, wherein the connection weights of the hidden layer and the output layer are matrices; wherein,
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811173901.5A CN109389207A (en) | 2018-10-09 | 2018-10-09 | A kind of adaptive neural network learning method and nerve network system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811173901.5A CN109389207A (en) | 2018-10-09 | 2018-10-09 | A kind of adaptive neural network learning method and nerve network system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109389207A true CN109389207A (en) | 2019-02-26 |
Family
ID=65426789
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811173901.5A Pending CN109389207A (en) | 2018-10-09 | 2018-10-09 | A kind of adaptive neural network learning method and nerve network system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109389207A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
-
2018
- 2018-10-09 CN CN201811173901.5A patent/CN109389207A/en active Pending
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US12020476B2 (en) | 2017-03-23 | 2024-06-25 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US12086097B2 (en) | 2017-07-24 | 2024-09-10 | Tesla, Inc. | Vector computational unit |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11797304B2 (en) | 2018-02-01 | 2023-10-24 | Tesla, Inc. | Instruction set architecture for a vector computational unit |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US12079723B2 (en) | 2018-07-26 | 2024-09-03 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11983630B2 (en) | 2018-09-03 | 2024-05-14 | Tesla, Inc. | Neural networks for embedded devices |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11908171B2 (en) | 2018-12-04 | 2024-02-20 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US12136030B2 (en) | 2018-12-27 | 2024-11-05 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US12014553B2 (en) | 2019-02-01 | 2024-06-18 | Tesla, Inc. | Predicting three-dimensional features for autonomous driving |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389207A (en) | A kind of adaptive neural network learning method and nerve network system | |
| CN108491765B (en) | Vegetable image classification and identification method and system | |
| US9619749B2 (en) | Neural network and method of neural network training | |
| CN107729999A (en) | Consider the deep neural network compression method of matrix correlation | |
| CN113469356A (en) | Improved VGG16 network pig identity recognition method based on transfer learning | |
| CN107798349B (en) | Transfer learning method based on depth sparse self-coding machine | |
| CN107689224A (en) | The deep neural network compression method of reasonable employment mask | |
| CN110222163A (en) | A kind of intelligent answer method and system merging CNN and two-way LSTM | |
| CN107506590A (en) | A kind of angiocardiopathy forecast model based on improvement depth belief network | |
| WO2015134900A1 (en) | Neural network and method of neural network training | |
| CN109492750B (en) | Zero sample image classification method based on convolutional neural network and factor space | |
| CN106777402A (en) | A kind of image retrieval text method based on sparse neural network | |
| CN111079837B (en) | Method for detecting, identifying and classifying two-dimensional gray level images | |
| CN110598737B (en) | Online learning method, device, equipment and medium of deep learning model | |
| CN115511069A (en) | Neural network training method, data processing method, device and storage medium | |
| CN113157919A (en) | Sentence text aspect level emotion classification method and system | |
| CN117174163A (en) | Virus evolution trend prediction method and system | |
| CN107528824A (en) | A kind of depth belief network intrusion detection method based on two-dimensionses rarefaction | |
| CN103559510B (en) | Method for recognizing social group behaviors through related topic model | |
| CN111461229A (en) | Deep neural network optimization and image classification method based on target transfer and line search | |
| CN110288002A (en) | A kind of image classification method based on sparse Orthogonal Neural Network | |
| Zhu et al. | Emotion Recognition in Learning Scenes Supported by Smart Classroom and Its Application. | |
| Hu et al. | Tree species identification based on the fusion of multiple deep learning models transfer learning | |
| CN109118483A (en) | A kind of label quality detection method and device | |
| CN115331045A (en) | Neural network adaptive expansion pruning-based visual object classification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190226 |
|
| RJ01 | Rejection of invention patent application after publication |