Functional Actions

How it works

Highlighted claims

• The action is defined through a log-quantile-density transformation of a kernel density estimate of daily steps. — Precision Physical Activity Prescription via Reinforcement Learning for Functional Actions
• Daily step distributions are represented as functional actions because participants' daily step trajectories are not temporally aligned. — Precision Physical Activity Prescription via Reinforcement Learning for Functional Actions
• The LQD representation enables B-spline policies by mapping density functions into an unconstrained function space. — Precision Physical Activity Prescription via Reinforcement Learning for Functional Actions
• The learned policy is back-transformed from LQD space into quantile or density space for interpretation. — Precision Physical Activity Prescription via Reinforcement Learning for Functional Actions