From: Michael Orlitzky <michael@orlitzky.com>
Date: Mon, 25 Mar 2013 16:49:58 +0000 (-0400)
Subject: Use the infinity norm in vanilla_cgm().
X-Git-Url: https://gitweb.michael.orlitzky.com/?a=commitdiff_plain;h=1894f0ed32e27eda86ad85065131151c17817895;p=octave.git

Use the infinity norm in vanilla_cgm().
Simplify the loop in vanilla_cgm().
---

diff --git a/optimization/vanilla_cgm.m b/optimization/vanilla_cgm.m
index 9ada2cb..c4d8f07 100644
--- a/optimization/vanilla_cgm.m
+++ b/optimization/vanilla_cgm.m
@@ -40,17 +40,11 @@ function [x, k] = vanilla_cgm(A, b, x0, tolerance, max_iterations)
   sqrt_n = floor(sqrt(length(x0)));
 
   k = 0;
-  xk = x0; % Eschew the 'k' suffix on 'x' for simplicity.
+  xk = x0;
   rk = A*xk - b; % The first residual must be computed the hard way.
   pk = -rk;
 
-  while (k <= max_iterations)
-    if (norm(rk) < tolerance)
-       % Success.
-       x = xk;
-       return;
-    end
-
+  while (k <= max_iterations && norm(rk, 'inf') > tolerance)
     alpha_k = step_length_cgm(rk, A, pk);
     x_next = xk + alpha_k*pk;