NgHanWei
commented
3 weeks ago Hi, meta-learning as a broad principle utilizes conventional learning and meta-learning followed by meta-testing as a whole. In the original meta-learning methodology, the model is first trained conventionally on a known task. Meta-learning is then applied to the model to allow it to learn parameters that let it adapt quickly to a new task. Meta-testing is simply the testing of the meta-trained model with a new task.
You may run train_motor_LOSO_all.py which trains a subject-independent meta-trained model.
Hello, taking motor imagery as an example, I would like to ask you which python files correspond to conventional learning, meta-learning and meta-testing in the article? And, does your article mean that the first step is conventional learning and then the second step is meta-learning? How does this look in code? Thank you very much!