where $Qc$ is the quantity of child $c$, $Q{tot}$ is the total quantity of the children and $CF_c$ is the cost function of the child. When a node has a large number of terms (greater than 1000), this cost function becomes large enough to potentially make the computation infeasible.
The solution is to introduce a shadow variable and constraint. If the variable is named $V$, the constraint is
$$\text{cost function} - V \perp V$$
This simplification greatly reduces memory usage and computation time.
Cost functions have the form
$$\sum_{c\in \text{children}} \frac{Qc}{Q{tot}}\cdot CF_c$$
where $Qc$ is the quantity of child $c$, $Q{tot}$ is the total quantity of the children and $CF_c$ is the cost function of the child. When a node has a large number of terms (greater than 1000), this cost function becomes large enough to potentially make the computation infeasible.
The solution is to introduce a shadow variable and constraint. If the variable is named $V$, the constraint is
$$\text{cost function} - V \perp V$$
This simplification greatly reduces memory usage and computation time.