Lab 5 2014
From 6.034 Wiki
(→Input(DifferentiableElement, ValuedElement)) |
|||
| Line 1: | Line 1: | ||
| - | __TOC__ | + | <!--__TOC__ |
=Hints= | =Hints= | ||
https://docs.google.com/View?id=dhqhm2bq_183ndrncvvr | https://docs.google.com/View?id=dhqhm2bq_183ndrncvvr | ||
| Line 432: | Line 432: | ||
<!-- | <!-- | ||
To make that a bit easier, we've made it possible for you to check your proposed ensemble on the machine we're actually testing on. Here is [http://samoa.csail.mit.edu:8080/boosted_orange?ensemble=maj+knn+svml+svmp3+svmr+svms+nb an example query] with all of the classifiers in learner except the decision tree. | To make that a bit easier, we've made it possible for you to check your proposed ensemble on the machine we're actually testing on. Here is [http://samoa.csail.mit.edu:8080/boosted_orange?ensemble=maj+knn+svml+svmp3+svmr+svms+nb an example query] with all of the classifiers in learner except the decision tree. | ||
| + | --> | ||
--> | --> | ||
Revision as of 22:24, 10 September 2011
Boosting and Orange
It turns out that different architectures yield different results from describe_and_classify on the vampire and H004 datasets. Some folks have perfectly reasonable answers to the short answer part given the output on their machine, but do not pass the tests with those answers. To try and standardize, we've put up the output from the two runs from which we generated the answers.
For the same reason, some folks are getting 74% on their home computers very easily. The question was intended to set a lower limit which was larger than any of the classifiers by itself. If you're getting 74, but the ensemble classifier test is failing, you'll need to try harder. You'll want to use some subset of the classifiers. Think about why the decision tree isn't a good choice as a base classifier for boosting.
-->
