Feedback for 3.8: Active Learning Applied

Clarity: 5/5
Quality: 4/5
Relevance: 5/5

Errors:
Suggestions:

– It might be good to give a warning that using a predefined data set can give a wrong impression about active learning. You are in fact not performing new experiments based on suggestions of the acquisition function. You are sequentially including the data points in your data set that have the largest value of the acquisition function. It is very likely that if you would present the acquisition function with the entire compositional space of interest, that the model training and data acquisition would occur entirely differently.

Location: AI for Materials Industry » Case 3: Optimizing glass composition using active learning » 3.8: Active Learning Applied

Clarity: 5/5
Quality: 4/5
Relevance: 5/5

Errors:
Suggestions:

– It might be good to give a warning that using a predefined data set can give a wrong impression about active learning. You are in fact not performing new experiments based on suggestions of the acquisition function. You are sequentially including the data points in your data set that have the largest value of the acquisition function. It is very likely that if you would present the acquisition function with the entire compositional space of interest, that the model training and data acquisition would occur entirely differently.

  • Actions:

Leave a Reply

Your email address will not be published. Required fields are marked *