def tune(self, tune_func, logger=None, type=max):
"""Tune values of parameters
Items in tuning grid are iteratively passed to `tune_func` and return
value of this function is used as criterion value.
It returns optimal value of criterion and points with associated
criterion value. If `points` is not None, it is used as default
dictionary for returned points. Points is a dictionary with parameter
values as keys and criterion value as values. There is a special key
'header' containing names of parameters.
"""
grid_keys, grid_values = self._getKeysValues()
crit = self._criterion
for i, values in enumerate(grid_values):
new_values = dict(zip(grid_keys, values))
old_crit = self.tunedValueFor(new_values)
if old_crit is not None:
crit[i] = old_crit
continue
if logger is not None:
self._logDict(logger, 'Setting values of parameters:',
new_values)
crit[i] = tune_func(**new_values)
if logger is not None:
logger.info("Criterion value: %s", strnumber(crit[i]))
logger.info("\n")
tuned_crit, tuned_values = self.optimum(type)
if logger is not None:
self._logTable(logger)
self._logTuned(logger, tuned_crit, tuned_values)
return tuned_crit, tuned_values
def tune(self, tune_func, logger=None, type=max):
"""Tune values of parameters
Items in tuning grid are iteratively passed to `tune_func` and return
value of this function is used as criterion value.
It returns optimal value of criterion and points with associated
criterion value. If `points` is not None, it is used as default
dictionary for returned points. Points is a dictionary with parameter
values as keys and criterion value as values. There is a special key
'header' containing names of parameters.
"""
grid_keys, grid_values = self._getKeysValues()
crit = self._criterion
for i, values in enumerate(grid_values):
new_values = dict(zip(grid_keys, values))
old_crit = self.oldCriterion(new_values)
if old_crit is not None:
crit[i] = old_crit
continue
if logger is not None:
self._logDict(logger, 'Setting values of parameters:', new_values)
crit[i] = tune_func(**new_values)
if logger is not None:
logger.info("Criterion value: %s", strnumber(crit[i]))
logger.info("\n")
tuned_crit, tuned_values = self.optimum(type)
if logger is not None:
self._logTable(logger)
self._logTuned(logger, tuned_crit, tuned_values)
return tuned_crit, tuned_values
def _logDict(self, logger, header, d):
logger.info(header)
logger.info("=" * len(header))
for key, val in d.iteritems():
logger.info(" %s: %s", key, strnumber(val))
def _logDict(self, logger, header, d):
logger.info(header)
logger.info("="*len(header))
for key, val in d.iteritems():
logger.info(" %s: %s", key, strnumber(val))
