Tricki
a repository of mathematical know-how

To optimize a sum try making the terms roughly equal in size

Stub iconThis article is a stub.This means that it cannot be considered to contain or lead to any mathematically interesting information.

Quick description

Suppose one wants to find the parameter \lambda that minimizes a quantity f(\lambda) + g(\lambda), where f is an increasing non-negative function in \lambda and g is a decreasing non-negative function in \lambda. One can of course use calculus methods to do this, by finding the value(s) of \lambda where the derivative of f(\lambda)+g(\lambda) vanishes. But a quick and dirty way to find the minimum approximately is just find the value \lambda_0 where the two functions agree:

 f(\lambda_0) = g(\lambda_0).

Indeed, since f(\lambda)+g(\lambda) \geq f(\lambda_0)+0 = g(\lambda_0) for \lambda \geq \lambda_0, and f(\lambda)+g(\lambda) \geq 0+g(\lambda_0) = g(\lambda_0) for \lambda \leq \lambda_0, we see that the minimal value of f(\lambda)+g(\lambda) lies between g(\lambda_0) and 2g(\lambda_0).

More generally, once one finds a \lambda_0 where f(\lambda_0) and g(\lambda_0) are comparable in magnitude, this is already enough to compute the minimum of f(\lambda)+g(\lambda) up to multiplicative constants.

Prerequisites

Calculus

Example 1

(Optimize expressions such as \lambda^\alpha A + \frac{B}{\lambda^\beta})

General discussion

When optimizing a sum f_1(\lambda) + \ldots + f_n(\lambda), where the intermediate f_j(\lambda) are somehow "in between" f_1(\lambda) and f_n(\lambda), the above heuristic is often effective (up to a factor of n, perhaps) if one looks for the \lambda which balances the two extreme terms f_1(\lambda) and f_n(\lambda). (Need an example of this...)

Comments

Post new comment

(Note: commenting is not possible on this snapshot.)

Before posting from this form, please consider whether it would be more appropriate to make an inline comment using the Turn commenting on link near the bottom of the window. (Simply click the link, move the cursor over the article, and click on the piece of text on which you want to comment.)

snapshot
Notifications