Profilers are java objects that function similar to the logging objects from java.util.logging. They allow you to record different statistics during the modeling process that you can use to track and assess the performance of the modeling algorithms used.
Examples of profilers are:
- Track the average time it takes to evaluate a sample
- Track the score of the current best model
- Track the value of the model parameters as they are optimized
- Track the memory used
A large number of profilers is included but only a small number are enabled by default (see FAQ: How do I enable more profilers for how to enable them). You can find them in the configuration file and their output in the
output/<run_name+date+time>/profilers directory. It is very easy to use and define your own. Look at the Profiler and AdaptiveModelBuilder.m sources to see how they are used.