__init__(self, features_dir, bin_dir, norm_file=None, tmp_dir='tmp', out_dir='./', features_ext='.mat', log_on_screen=False, log_file=None, executor=None)
(Constructor)

The Experiment object constructor.

• features_dir: (string) directory containing the features data
• bin_dir: (string) directory containing the Torch executable
• norm_file: (string) file containing the world model file list (default None)
• tmp_dir: (string) name of the temporary directory (default "tmp")
• out_dir: (string) output directory to put models, scores, results (default "./")
• features_ext: (string) features files extension (default ".mat")
• log_on_screen: (boolean) put the log message on stdout/stdin (default False)
• log_file: (string) file to put the log messages (default None)
• executor: (pyVerif.Executor) instance: object to run external command (default None)

build_clients_scores(self, for_dev)

Compute the clients scores for all the models.

Specify for_dev=True if you want to compute for dev set, and for_dev=False if you want to compute for test set.

• for_dev: (boolean) select if you want to work on the development set or on the test set

load_protocols(self, base_dir, variant)

Load both dev and test protocols.

Load both protocols dev and test from the given directory and the wanted variant.

• base_dir: (string) directory containing the protocol files
• variant: (string) variant of the protocol, for example, fem, nist, P, ...

log(self, txt)

log_options(self, options, text=None)

Log the given options.

• options: the options to log, of type Values, as returned by OptionParser.parse_args()
• text: optional text which will be printed before the options

merge_score(self, out_file, for_dev)

Merge the computed into a result file.

Merge the score for either dev or test, putting the result in out_file.

This method should be overwritted in the new classes, as it does nothing by default.

• for_dev: (boolean) select if you want to work on the development set or on the test set

set_cmds(self, train_cmd, train_opts, test_cmd, test_opts)

Set the commands for training and testing the clients models.

• train_cmd: (string) command for trainning the models
• train_opts: (string) command line options for trainning the models
• test_cmd: (string) command for computing scores
• test_opts: (string) command line options for computing scores

set_dirs(self, models_dir, scores_dir)

train_clients_models(self, for_dev)

Train all the clients models.

This method should be overwritted in the new classes, as it does nothing by default.

• for_dev: (boolean) select if you want to work on the development set or on the test set