Suppose we have some functions $f_1(x), f_2(x), \ldots, f_n(x)$ with $x \in \mathbb{Z}^n$.

We can denote the subset $X_1$ of $\mathbb{Z}^n$ that maximizes $f_1(x)$ as:

$$X_1 = \underset{x \in \mathbb{Z}^n}{\operatorname{arg\,max}}\, f_1(x)$$

Now, suppose there is a kind of "priority" in which I also want to maximize $f_2$, as long as I keep maximizing $f_1$. This could be represented as:

$$X_2 = \underset{x \in X_1}{\operatorname{arg\,max}}\, f_2(x)$$

The same for $f_3$:

$$X_3 = \underset{x \in X_2}{\operatorname{arg\,max}}\, f_3(x)$$

So on and so forth...

Is there some, more concise, notation to represent this "maximization priority"?


I figured it out and decided to post an answer in case more people are interested.

This maximization with priority is indeed a type of multi-objective optimization. More specific it is what is called the lexicographic method. In the lexicographic method, functions are arranged in priority. The formal definition is as follow:

$$ \begin{aligned} & \underset{\mathbf{x} \in \mathbf{X}}{\text{minimize}} & & f_l(\mathbf{x}) \\ & \text{subject to} & & f_j(\mathbf{x})\leq\mathbf{y}^*_j,\;j=1,\dotsc,l-1,\\ \end{aligned} $$

where $\mathbf{y}^*_j$ is the optimum of the $j$th objective function with $l=j$.

Now going back to the question, we may adapt the lexicographic optimization to fit in the "arg max" problem, it becomes:

$$ \begin{aligned} & \underset{x \in \mathbb{Z}^n}{\operatorname{arg\,max}} & & f_l(x) \\ & \text{subject to} & & f_j(x)\leq\mathbf{y}^*_j,\;j=1,\dotsc,l-1,\\ \end{aligned} $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.