The add listener also allows to add a listener after the evaluation
is already started. I believe we don't want (and allow) people to
do that, since the evaluation should always look at all available samples.
First it could be easily abused, e.g add a listener from a different thread
while the evaluation is running, or when a listener is added later it might
not be valid for the cross validation case, where we need to somehow
pass down the information about the current fold.
What happens if a listener is added while fold number three is evaluated?
Maybe the term listener is a bit misleading here, in the eclipse APIs I
believe
such "listeners" are frequently called monitors.
Jörn
On 8/18/11 10:59 PM, Jörn Kottmann wrote:
On 8/18/11 10:45 PM, [email protected] wrote:
We should add more listeners. For example one to get detailed F-measure,
another to have detailed output of false positives and false negatives.
First I tried passing a list of listeners, but it would require more
code.
Why is that? You could just use these java 5 variable length arguments.
The removeListener implies to people that they have to remove their
listeners again after they are done using the evaluator, but this should
not be necessary. After the results are retrieved everything will be
thrown
away.
Jörn
