diff options
| -rw-r--r-- | lib/utils.py | 62 |
1 files changed, 46 insertions, 16 deletions
diff --git a/lib/utils.py b/lib/utils.py index b496a7b..e9e347e 100644 --- a/lib/utils.py +++ b/lib/utils.py @@ -3,6 +3,7 @@ import numpy as np arg_support_enabled = True def is_numeric(n): + """Check if n is numeric (i.e., can be converted to int).""" if n == None: return False try: @@ -25,15 +26,44 @@ def param_slice_eq(a, b, index): example: ('foo', [1, 4]), ('foo', [2, 4]), 0 -> True ('foo', [1, 4]), ('foo', [2, 4]), 1 -> False + """ if (*a[1][:index], *a[1][index+1:]) == (*b[1][:index], *b[1][index+1:]) and a[0] == b[0]: return True return False -def compute_param_statistics(by_name, by_param, parameter_names, num_args, state_or_trans, key): +def compute_param_statistics(by_name, by_param, parameter_names, arg_count, state_or_trans, attribute): + """ + Compute standard deviation and correlation coefficient for various data partitions. + + arguments: + by_name -- ground truth partitioned by state/transition name. + by_name[state_or_trans][attribute] must be a list or 1-D numpy array. + by_param -- ground truth partitioned by state/transition name and parameters. + by_name[(state_or_trans, *)][attribute] must be a list or 1-D numpy array. + parameter_names -- list of parameter names, must have the same order as the parameter + values in by_param (lexical sorting is recommended). + arg_count -- dict providing the number of functions args ("local parameters") for each function. + state_or_trans -- state or transition name, e.g. 'send' or 'TX' + attribute -- model attribute, e.g. 'power' or 'duration' + + returns a dict with the following content: + std_static -- static parameter-unaware model error: stddev of by_name[state_or_trans][attribute] + std_param_lut -- static parameter-aware model error: mean stddev of by_param[(state_or_trans, *)][attribute] + std_by_param -- static parameter-aware model error ignoring a single parameter. + dictionary with one key per parameter. The value is the mean stddev + of measurements where all other parameters are fixed and the parameter + in question is variable. E.g. std_by_param['X'] is the mean stddev of + by_param[(state_or_trans, (X=*, Y=..., Z=...))][attribute]. + std_by_arg -- same, but ignoring a single function argument + Only set if state_or_trans appears in arg_count, empty dict otherwise. + corr_by_param -- correlation coefficient + corr_by_arg -- same, but ignoring a single function argument + Only set if state_or_trans appears in arg_count, empty dict otherwise. + """ ret = { - 'std_static' : np.std(by_name[state_or_trans][key]), - 'std_param_lut' : np.mean([np.std(by_param[x][key]) for x in by_param.keys() if x[0] == state_or_trans]), + 'std_static' : np.std(by_name[state_or_trans][attribute]), + 'std_param_lut' : np.mean([np.std(by_param[x][attribute]) for x in by_param.keys() if x[0] == state_or_trans]), 'std_by_param' : {}, 'std_by_arg' : [], 'corr_by_param' : {}, @@ -41,27 +71,27 @@ def compute_param_statistics(by_name, by_param, parameter_names, num_args, state } for param_idx, param in enumerate(parameter_names): - ret['std_by_param'][param] = _mean_std_by_param(by_param, state_or_trans, key, param_idx) - ret['corr_by_param'][param] = _corr_by_param(by_name, state_or_trans, key, param_idx) - if arg_support_enabled and state_or_trans in num_args: - for arg_index in range(num_args[state_or_trans]): - ret['std_by_arg'].append(_mean_std_by_param(by_param, state_or_trans, key, len(parameter_names) + arg_index)) - ret['corr_by_arg'].append(_corr_by_param(by_name, state_or_trans, key, len(parameter_names) + arg_index)) + ret['std_by_param'][param] = _mean_std_by_param(by_param, state_or_trans, attribute, param_idx) + ret['corr_by_param'][param] = _corr_by_param(by_name, state_or_trans, attribute, param_idx) + if arg_support_enabled and state_or_trans in arg_count: + for arg_index in range(arg_count[state_or_trans]): + ret['std_by_arg'].append(_mean_std_by_param(by_param, state_or_trans, attribute, len(parameter_names) + arg_index)) + ret['corr_by_arg'].append(_corr_by_param(by_name, state_or_trans, attribute, len(parameter_names) + arg_index)) return ret -def _mean_std_by_param(by_param, state_or_tran, key, param_index): +def _mean_std_by_param(by_param, state_or_tran, attribute, param_index): u""" Calculate the mean standard deviation for a static model where all parameters but param_index are constant. arguments: by_param -- measurements sorted by key/transition name and parameter values state_or_tran -- state or transition name (-> by_param[(state_or_tran, *)]) - key -- model attribute, e.g. 'power' or 'duration' - (-> by_param[(state_or_tran, *)][key]) + attribute -- model attribute, e.g. 'power' or 'duration' + (-> by_param[(state_or_tran, *)][attribute]) param_index -- index of variable parameter - Returns the mean standard deviation of all measurements of 'key' + Returns the mean standard deviation of all measurements of 'attribute' (e.g. power consumption or timeout) for state/transition 'state_or_tran' where parameter 'param_index' is dynamic and all other parameters are fixed. I.e., if parameters are a, b, c ∈ {1,2,3} and 'index' corresponds to b, then @@ -73,18 +103,18 @@ def _mean_std_by_param(by_param, state_or_tran, key, param_index): param_partition = [] for k, v in by_param.items(): if param_slice_eq(k, param_value, param_index): - param_partition.extend(v[key]) + param_partition.extend(v[attribute]) if len(param_partition): partitions.append(param_partition) else: print('[W] parameter value partition for {} is empty'.format(param_value)) return np.mean([np.std(partition) for partition in partitions]) -def _corr_by_param(by_name, state_or_trans, key, param_index): +def _corr_by_param(by_name, state_or_trans, attribute, param_index): if _all_params_are_numeric(by_name[state_or_trans], param_index): param_values = np.array(list((map(lambda x: x[param_index], by_name[state_or_trans]['param'])))) try: - return np.corrcoef(by_name[state_or_trans][key], param_values)[0, 1] + return np.corrcoef(by_name[state_or_trans][attribute], param_values)[0, 1] except FloatingPointError as fpe: # Typically happens when all parameter values are identical. # Building a correlation coefficient is pointless in this case |