Detour D — Optimization, minimum and argmin (detailed)
On this page 3 headings
An objective is a number we want to make small (or large). A minimum is the smallest attained objective VALUE. An argmin is the INPUT, or set of inputs, where that value is attained. A supremum is the least upper bound and need not be attained; we will explore it in detail in Part 2.
Minimum versus argmin
Take
Since every square is nonnegative, .
Equality holds only when , so .
Therefore
whereas
For , find the minimum value and argmin.
Show solution
minimum 7; argmin .
The optimizer versus the optimum
Example A:
The square is at least zero, so at ;
Example B:
The minimum is 0, achieved at and ;
One objective can have more than one minimizing input.
Common mistake
Writing
for Example A confuses an input with an objective value. Another mistake is to assume that writing an optimization problem means a training algorithm has already solved it. The example states the goal; finding a practical algorithm is a separate question.
