4/2/2023 0 Comments Airflow technologyIf the value of the error after the next training becomes much higher or lower we should not use this model in production. The idea is that the researchers know the values of the error metrics for the training and test sets on the final step of the prototyping. The simplest what we can do - calculate the metrics using the training dataset. But we still have to monitor the quality of the re-trained model. In reality, it can be hard and expensive to keep the representative test dataset.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |