EMA (standing for EMpirical Analysis of algorithms) is a tool created for providing an effortless way to assess the performance of algorithms. EMA may be handy in many scenarios, including:

CHANGE LOG:
2.5:
 Fixed standard deviation calculation (divisor from 'N' to 'N1');
 Outlier detection and configurable removal of mean and standard deviation calculations;
 Added the parameter 'minimum number of samples';
 Convergence Factor can now relate to any resource, not only "Time"
 More function formats added to the fittness procedure.
 More initial values in distinct order of magnitudes (to offer better chances of producing a better fitting).
 In shared computers for running EMA, there will be a maximum number of instances of EMA allowed to concurrently run,
given by max{1,cores/2}.
 More customizations for graphs:
 graph size, line thickness, colors, font family, and customized captions in keys;
 numbers like '1.2e09' now appear as '1.2x10^9'
 plot data within a specific x and/or y range
 application of factors to customized resources values (e.g., to plot them as real values instead of integers)
 encodings for labels
2.4:
 Added support for outputting graphs in eps format.
 Old graph files are backed up whiile outputting new graphs.
 Added support for positioning the key on graphs and changing font family/size on them
 Fixed issues related to running EMA in Windows.
 Reporting execution time on screen with less frequency as the execution time gets longer
2.3:
 Added independent term in a monomial so that the fitting can be more accurate.
 Used allocated memory instead of resident memory as 'used memory' for counting memory consumption (Linux only).
 Fixed terminology (Equation ==> Function)
 The report of functions after estimation can be now requested to be printed out in the console through a
parameter (and the report layout has been improved a little bit).
 Cosmetic changes in the chart layout (colors, line thickness, points, headers).
 getResourceUsageFunction() now can set the error threshold so that a function is considered equivalent to that
of minimumerror.