When to Sell an Asset? - A Distribution Builder Approach

What the math gives to ML

The paper replaces scalar utility maximization with a target-distribution objective: choose a stopping time whose stopped diffusion has a prescribed law. Its transferable asset is the coupling between path-dependent stopping rules and terminal distribution matching, rather than the finance-specific asset-sale interpretation. A practical neural adaptation is to learn a differentiable stopping hazard along a stochastic latent trajectory, with a distributional loss enforcing the desired terminal law; this can provide adaptive-compute diffusion sampling or distribution-constrained RL policies.