DFQI transfers knowledge for offline Q-learning
The method integrates source data via L1 penalty in target Q-function updates to improve policy learning with limited target data in marketing and healthcare.
Stories per day · 30 days
1 stories, newest first · one card per story