When artificial intelligence is used to train the model, data is the most important factor. Only through big data training can the trained model be closer to the actual demand. For example, the model training of autonomous driving is basically based on the self-driving car on-road test to collect data training models. Only cars can collect more possible scenes to ensure that the trained model is closer to the actual road conditions. There are two problems in AI training: First, how is the amount of data insufficient to train the model? Second, how large is the amount of data?
Zhao Zhihong, Product Marketing Manager, MathWorks
Migration Learning: Train the neural network through the model generation <br style="padding: 0px; margin: 0px;"> If there is no data to do artificial intelligence? This is the confusion of many startups. In the author's concept, the normal AI development process is to collect data and then train the model. At the recent Matlab conference, Mr. Zhao Zhihong, product marketing manager at MathWorks, gave a different answer. He said, “Data, output and models are a step in the entire AI development process. If developers want to do artificial intelligence, there is no Too much data can be used in migration learning.†What is the concept of migration learning? Do not understand from the actual examples. For example, the artificial intelligence method is used to predict the fault of the wind turbine. The user cannot wait until the large amount of fault data is collected, and then predicts. This does not meet the goal of equipment maintenance. Instead, the model of the wind turbine generator is first established by using the Simul ink model. By correcting this model very close to the wind turbine, and then generating fault data through the model, and then training the neural network of machine learning or deep learning.
Mr. Zhao Zhihong said, “The advantage of model-based design is that it builds a model before actually making a product. This model is very close to the actual product, both from numerical calculations, behavioral calculations, and real results. Very close. Many engineers have already built this model when designing the product. It is easy to generate fault data when adding fault conditions to this model, which is much easier than the faults generated on actual equipment. Therefore, faults can be made. Forecasting and maintenance."
Labeling big data through deep learning <br style="padding: 0px; margin: 0px;"> With the large number of sensors, the amount of data has entered the era of flooding, making it difficult to feature signatures for artificial intelligence. Deep learning is marked. Mr. Zhao Zhihong explained, “The characteristic of deep learning is that it does not need to manually find the eigenvalues ​​manually, and the system can automatically extract the eigenvalues ​​from the data. There are many auto-marking tools and functions in MATLAB. LiDAR 3D point cloud technology can be used for every point. Marking, clustering the points together into a target model, and then distinguishing the actual objects represented by the target. Some users have developed using MATLAB tools, and the famous auto parts company AUTOLIV is using this method. Automatic data annotation."
Taking the intelligent detection of puffed food as an example, researchers can extract features when users bite food, and measure the crispness of food with bite sound and bite strength. With these two characteristics, it is also necessary to develop a machine learning classifier. MATLAB provides a classification learner. Developers don't need to try out the algorithms of various classifiers through this tool, use the APP provided by MATLAB to try all the algorithms at once. After starting the app, the user selects the data and the classifier that needs to be trained, and then trains. During the training process, the user can see the overall results of each classifier, select the one with the highest accuracy, and then conduct more investigations and research. If the user has not studied AI, you can use the APP provided by MATLAB to learn and try all the algorithms for machine learning.
Is it difficult to do deep learning in MATLAB? In the words of Mr. Zhao Zhihong, you only need to write 5 lines of MATLAB code, and users can build a network that can identify food. I haven't been exposed to MATLAB for many years, so I have no research, and engineers can try if they need it.
How does it work with open source code?
Now that open source is very hot on the market, engineers can quickly get the code they need in the open source community. There are also many open source resources in the field of artificial intelligence. As a closed resource, MATLAB and open source code may be slower than update code, and In the charging mode, engineers are also worried about the integration of the two. When the author asked this question, Mr. Zhao Zhihong explained, “There are a lot of development teams behind open source code. They volunteer to develop. We also have our own development team. We don’t define open source code and MATLAB as a competitive relationship. In fact, we have their own characteristics and strengths, sometimes we can be shared with the open source code. there are some open source code can be run inside or directly in MATLAB calls, we jointly contribute to scientific and technological progress. Although we are paid software, but provides not only tools, there are a lot of service. "
"When users encounter problems using MATLAB, we followed a strong technical support team to help users solve these problems. In addition, our tool not only produce correct results for a certain function, we also consider this feature has been very easy to use, such as for AI applications, we have been very appropriate for the interface area of expertise, but both let users feel interface using the familiar MATLAB or Simul ink of which is to our advantage, if the user or Simul ink in MATLAB add new features, learning threshold is not high, easy to use design. in addition to providing tools, we returned to the user to provide a lot of convenience, it is open source tools from those things are not. "Mr. Zhao Zhihong supplement.
4Lan Software Router,4Lan Mini Pc,4Lan Firewall Router,4 Ethernet Mini Pc
Shenzhen Innovative Cloud Computer Co., Ltd. , https://www.xcypc.com