Gilles Sadowski created MATH-1553:
-------------------------------------
Summary: Fixed seed hides sensitivity of a unit test.
Key: MATH-1553
URL: https://issues.apache.org/jira/browse/MATH-1553
Project: Commons Math
Issue Type: Test
Affects Versions: 3.6.1
Reporter: Gilles Sadowski
Fix For: 4.X
In the tests suite originally provided to exercise {{BOBYQAOptimizer}}, a
[line|https://gitbox.apache.org/repos/asf?p=commons-math.git;a=blob;f=src/test/java/org/apache/commons/math4/optim/nonlinear/scalar/noderiv/OptimTestUtils.java;h=df6cd1f1dad0571a781f2ffdc72d2024c702541c;hb=2470c3ff28d5e5f378cab14838fced2b2643701e#l348]
seems intended to ensure some kind of robustness of the optimizer. However,
changing the seed value can make the [corresponding unit
test|https://gitbox.apache.org/repos/asf?p=commons-math.git;a=blob;f=src/test/java/org/apache/commons/math4/optim/nonlinear/scalar/noderiv/BOBYQAOptimizerTest.java;h=6aae6a3682095ece19019c7797b370d0fe279577;hb=HEAD#l118]
fail.
This is even more blatant for [this
test|https://gitbox.apache.org/repos/asf?p=commons-math.git;a=blob;f=src/test/java/org/apache/commons/math4/optim/nonlinear/scalar/noderiv/SimplexOptimizerMultiDirectionalTest.java;h=cf108df6e20b0498c2f28b1d9248a996cd2e014c;hb=HEAD#l216]
of {{MultiDirectionalSimplex}} where the number of evaluations required to
reach the solution heavily depends on the seed value (from the order of a
thousand up to several tens of millions).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
