[
https://issues.apache.org/jira/browse/MATH-1433?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16191501#comment-16191501
]
Stephen Crowley commented on MATH-1433:
---------------------------------------
Sure thing.. I'll try to get to it soon. It was a bug in my random
vector guess generating code that just took a while to track down. The
code as is, just ignores it as a "bad guess" and ends up returning the
initial starting point results and only throws the exception if the
initial starting point isn't there. I'm thinking something like an
exception in the initial guess should go ahead and cause the entire
optimize() call to fail since
one would think that the caller always wants to know something is messed
up first.
> BaseMultiStartMultivariateOptimizer.doOptimize() swallows exceptions
> --------------------------------------------------------------------
>
> Key: MATH-1433
> URL: https://issues.apache.org/jira/browse/MATH-1433
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 3.6
> Environment: // Multi-start loop.
> for (int i = 0; i < starts; i++) {
> // CHECKSTYLE: stop IllegalCatch
> try {
> // Decrease number of allowed evaluations.
> optimData[maxEvalIndex] = new MaxEval(maxEval -
> totalEvaluations);
> // New start value.
> double[] s = null;
> if (i == 0) {
> s = startPoint;
> } else {
> int attempts = 0;
> while (s == null) {
> if (attempts++ >= getMaxEvaluations()) {
> throw new
> TooManyEvaluationsException(getMaxEvaluations());
> }
> s = generator.nextVector();
> for (int k = 0; s != null && k < s.length; ++k) {
> if ((min != null && s[k] < min[k]) || (max !=
> null && s[k] > max[k])) {
> // reject the vector
> s = null;
> }
> }
> }
> }
> optimData[initialGuessIndex] = new InitialGuess(s);
> // Optimize.
> final PAIR result = optimizer.optimize(optimData);
> store(result);
> } catch (RuntimeException mue) {
> lastException = mue;
> }
> // CHECKSTYLE: resume IllegalCatch
> totalEvaluations += optimizer.getEvaluations();
> }
> Reporter: Stephen Crowley
> Labels: easyfix
>
> it took me a while to figure out that the random vector generator was
> throwing an exception that was being swallowed. that should never happen
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
