# Penalty Methods for Optimization

Consider the following problem: $$\text{min} \ f(x) \\ \text{subject to} \ h(x) = 0$$

This is a constrained optimization problem. We want to convert it into an unconstrained optimization problem so that we can use Newton's Method, quasi-Newton methods, etc..

So we have $$(1) \ \ \text{min} \ f(x) + \frac{1}{2} \rho \sum_{i=1}^{m} (g_{i}(x))^2 \\ (2) \ \ \text{min} \ f(x) + \rho \sum_{i=1}^{m} |g_{i}(x)|$$

Why is (2) exact and (1) not exact?

-
Ok, what exactly is $g_i(x)$ and where did $h(x)$ go? –  TenaliRaman May 4 '12 at 20:22