{"id":"arxiv/2611.00929","url":"https://arxiv.org/abs/2611.00929","tags":["stat.ME","math.ST"],"score":0,"title":"Deformed Q-learning for High-Dimensional Dynamic Treatment Regimes","domain":"stat.ME","source":"arxiv","authors":["Ming Cheng","Zhao Chen","Christina Dan Wang"],"reviews":0,"subject":"stat.ME","summary":"In high-dimensional dynamic treatment regimes, nonlinear terminal outcome models complicate upstream conditional-mean modeling and inference based on recursively estimated continuation values. We introduce a Deformed Q-learning procedure for general action spaces that accommodates a terminal-stage GLM, random designs, conditional heteroscedasticity, and fine-grained action discretization. The procedure changes the recursive criterion by propagating the terminal deformed optimal Q-function through sparse block-wise linear conditional-mean models. Under the stated models and regularity…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-31","reactions":{}}