interpolate_time_and_error

post_processors.interpolate_time_and_error(err_name='errors', time_name='timings', k=0, logtime=False, npoints=500, adaptive_grid=False, alg_name='algorithm')

Interpolates several error curves so that they all lie on the same grid. There is one grid per algorithm name by default. To use with caution as this can significantly bend the error curves.

Parameters:

• df (pandas dataframe) – raw shootout results dataframe with errors and time in lists.

• k (int, default 0) – the max time for interpolation will be the (longest - k) runtime for the algorithms. Set to k>0 if k runs are abnormally long.

• logtime (boolean, default False) – choose if interpolation grid is linear (False) or logarithmic (True). Set to True when timings are very different between several runs.

• npoints (int, default 500) – number of iterpolation points.

• adaptive_grid (bool, default False) – determines if each test condition has its own time grid. If True, a dirty BAD hack is used: shootout runs samples in most intern loop, therefore we can compute grids for each block of rows in df cut according to periodicity of seed.

• alg_name (string, default "algorithm") – the string value of the key in df that contains the algorithm name.