a forecasting problem for future prices.

 a high-dimensional stochastic control problem.

a static portfolio selection problem. 

a sequential decision-making problem under uncertainty.

can be integrated via penalties or architectural design. 

 reflect regulatory and economic requirements.

are secondary to the objective and can be ignored during training.

influence the feasible strategy space.

requires post-processing adjustments only.

influences the risk-return trade-off.

shapes the feasible region explored during training.

can be handled via penalty terms in the objective. 

It attempts to forecast market prices as accurately as possible.

It solves high-dimensional stochastic optimisation problems under constraints. 

It replaces management responsibility with automated decision-making.

It adapts balance sheet decisions to uncertainty rather than predicting the future.

equity is a derived (residual) quantity.

each time step can be optimised independently without affecting others.

funding decisions today affect tomorrow’s risk profile.

assets and liabilities evolve interdependently over time.