Conservative Continuous-Time Treatment Optimization
This addresses treatment optimization for patients with irregular medical data, though it appears incremental as it builds on existing stochastic control methods with a new regularizer.
The paper tackles the problem of treatment optimization from irregularly sampled patient trajectories by developing a conservative continuous-time stochastic control framework that adds a signature-based MMD regularizer to penalize out-of-support controls, with experiments showing improved robustness and performance compared to baselines.
We develop a conservative continuous-time stochastic control framework for treatment optimization from irregularly sampled patient trajectories. The unknown patient dynamics are modeled as a controlled stochastic differential equation with treatment as a continuous-time control. Naive model-based optimization can exploit model errors and propose out-of-support controls, so optimizing the estimated dynamics may not optimize the true dynamics. To limit extrapolation, we add a consistent signature-based MMD regularizer on path space that penalizes treatment plans whose induced trajectory distribution deviates from observed trajectories. The resulting objective minimizes a computable upper bound on the true cost. Experiments on benchmark datasets show improved robustness and performance compared to non-conservative baselines.