Difference: FJPPLAccelerateurRD (23 vs. 24)
Revision 242020-09-25 - Main.IsabelleRippBaudot
| Line: 1 to 1 | ||||||||
|---|---|---|---|---|---|---|---|---|
| ||||||||
| Line: 165 to 165 | ||||||||
| Added: | ||||||||
| > > | The goal of our project is to investigate and demonstrate of the use of Machine Learning (ML) techniques for advanced control and performance optimization of the accelerators and in particular the KEK injector. We search for very precise control and stability of the beams, better understanding of multi-parameter non-linear system with profound feature importance analysis. There are tree major steps in this project. One is the collection, processing, alignment, understanding and labelling of the raw data (machine parameters, diagnostics, BPMs etc, temperatures) to form the dataset for further deep learning. The next step is the development of the ML models, training, test and validation of several architectures of deep neural networks (DNNs) and convolutional neural networks (CNNs). Understanding of the models robustness with respect to the noise of different origins is crucial. The final step is tests of the models on the live data from the machine and analysis of the improvements of the Linac performance. The predicted parameters then could be inserted to the EPICS data channel for the monitoring purposes. | |||||||
|
French members: V. Kubytskyi, H. Guler, I. Chaikovska | ||||||||
View topic | History: r24 < r23 < r22 < r21 | More topic actions...
Copyright &© by the contributing authors. All material on this collaboration platform is the property of the contributing authors. Ideas, requests, problems regarding TWiki? Send feedback