There may be circumstances when nonlinear nutritional constraints are desirable. For instance, putting a lower limit on the energy density of the whole diet is a nonlinear constraint, when it is defined as a ratio: the mathematical expression of this constraint, which is nonlinear,
is: (X1.E1 + X2.E2 + X n.En) / (X1 + X2 + … Xn) > ED
where E1… En is the energy content for 100 g of food 1 to n and ED is the desired energy density. Optimization of these nonlinear models often cannot be achieved by solving simple equations or algorithms. Solutions are instead found by an iterative approach, creating a risk that a local optimum is selected instead of an overall general optimal solution. Hence, nonlinear functions should be avoided in the first instance. To achieve this, nonlinear
constraints should be reformulated into linear constraints, where possible. For example, the constraint on energy density can be expressed as a linear constraint as
follows: X1(E1 – ED) + X2(E2 – ED) + X n(En –ED) 0
These inequalities are linear, and therefore appropriate for a linear programming model. A similar transformation has been described previously for transforming the nonlinear phytate:zinc molar ratio constraint into a linear constraint (8). Keeping the phytate:zinc ratio below a
predetermined level may be needed to ensure a reasonable degree of zinc absorption (14).