1. Fine-tuning 微调 Transfer learning
Fine-tuning fine-tuning is the most important skill in the entire deep learning field. If the previous knowledge can only choose one, it is fine-tuning. Transfer learning Transfer learning is that you can use the parameters of other successful networks to train a model similar to you. Simply put, reuse.
For detailed knowledge, please refer to: Transfer learning Transfer Learning Guide
2. Code implementation
3. Q&A
-
- Fine tuning and transfer training are actually the same thing
-
- The source data set and the target data set are preferably similar, and the fine tuning effect is good.
-
- The auto-gluon automation framework mainly uses fine-tuning.
reference
https://www.bilibili.com/video/BV1Sb4y1d7CR?p=1