def main():
"""Main"""
parser = argparse.ArgumentParser()
parser.add_argument(
"-analyze_results_only",
action="store_true",
help="only analyze results, instead of generating them as well"
" (useful when results already generated")
parser.add_argument("-type",
choices=[
INSTANCE_COUNTS, FEATURE_COUNTS, NOISE_LEVELS,
SHUFFLING_COUNTS
],
default=INSTANCE_COUNTS)
parser.add_argument("-output_dir", required=True)
parser.add_argument("-seed", type=int, required=True)
args, pass_arglist = parser.parse_known_args(
) # Any unknown options will be passed to simulation.py
args.pass_arglist = " ".join(pass_arglist)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
args.logger = utils.get_logger(__name__,
"%s/run_simulations.log" % args.output_dir)
args.logger.info("Begin running/analyzing simulations")
args.config = load_config(args)
simulations = parametrize_simulations(args)
run_simulations(args, simulations)
analyze_simulations(args, simulations)
args.logger.info("End running simulations")
def main():
"""Main"""
parser = argparse.ArgumentParser()
parser.add_argument("-num_trials", type=int, default=3)
parser.add_argument("-start_seed", type=int, default=100000)
parser.add_argument("-type",
choices=[
constants.INSTANCE_COUNTS,
constants.FEATURE_COUNTS, constants.NOISE_LEVELS,
constants.SHUFFLING_COUNTS
],
default=constants.INSTANCE_COUNTS)
parser.add_argument(
"-summarize_only",
help="enable to assume that the results are already generated,"
" and just summarize them",
action="store_true")
parser.add_argument("-output_dir", required=True)
args, pass_arglist = parser.parse_known_args()
args.pass_arglist = " ".join(pass_arglist)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
args.logger = utils.get_logger(__name__,
"%s/run_trials.log" % args.output_dir)
trials = gen_trials(args)
run_trials(args, trials)
summarize_trials(args, trials)
def main():
"""Main"""
parser = argparse.ArgumentParser()
parser.add_argument("args_filename", help="pickle file containing arguments"
" passed by master.py")
cargs = parser.parse_args()
with open(cargs.args_filename, "rb") as args_file:
args = pickle.load(args_file)
logger = utils.get_logger(__name__, "%s/worker_%d.log" % (args.output_dir, args.task_idx))
pipeline(args, logger)
def main():
"""Main"""
parser = argparse.ArgumentParser()
parser.add_argument("-output_dir", help="name of output directory")
parser.add_argument(
"-dependence_assumption",
help=
"choice of dependence assumption used by Lynch and Guo (2016) procedure",
choices=[constants.POSITIVE, constants.ARBITRARY],
default=constants.POSITIVE)
parser.add_argument("-alpha", type=float, default=0.05)
parser.add_argument("-procedure",
default=constants.YEKUTIELI,
choices=[constants.YEKUTIELI, constants.LYNCH_GUO])
parser.add_argument(
"csv_filename",
help=
"CSV (with header) representing hierarchy, each row corresponding to one node:"
" the name of the node, the name of its parent node, the node's p-value and optionally a description of the node"
)
parser.add_argument("-effect_name", default="Effect size")
parser.add_argument(
"-tree_effect_size_threshold",
help="while generating output tree of rejected hypotheses,"
" only show nodes within given threshold of root (i.e. all nodes erased) effect size",
type=float,
default=1)
parser.add_argument("-color_scheme",
default="ylorrd9",
help="color scheme to use for shading nodes")
parser.add_argument("-color_range",
help="range for chosen color scheme",
nargs=2,
type=int,
default=[1, 9])
parser.add_argument(
"-sorting_param",
help="parameter to sort on for color grading",
default=constants.ADJUSTED_PVALUE,
choices=[constants.ADJUSTED_PVALUE, constants.EFFECT_SIZE])
parser.add_argument(
"-minimal_labels",
help="do not write descriptions/effect sizes on node labels",
action="store_true")
parser.add_argument(
"-rectangle_leaves",
help=
"enable to generate rectangular nodes for leaves of original hierarchy",
action="store_true")
args = parser.parse_args()
if not args.output_dir:
args.output_dir = "%s_alpha_%f_effect_threshold_%f" % (
os.path.splitext(os.path.basename(args.csv_filename))[0],
args.alpha, args.tree_effect_size_threshold)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
logger = utils.get_logger(
__name__, "%s/hierarchical_fdr_control.log" % args.output_dir)
logger.info("Begin hierarchical_fdr_control")
tree = build_tree(args, logger)
F, M = process_tree(logger, tree)
hierarchical_fdr_control(args, logger, F, M)
write_outputs(args, logger, tree)
logger.info("End hierarchical_fdr_control")
def main():
"""Main"""
parser = argparse.ArgumentParser()
parser.add_argument("-seed", type=int, default=constants.SEED)
parser.add_argument("-num_instances", type=int, default=10000)
parser.add_argument("-num_features", type=int, default=100)
parser.add_argument("-output_dir", help="Name of output directory")
parser.add_argument("-fraction_relevant_features", type=float, default=.05)
parser.add_argument(
"-noise_multiplier",
type=float,
default=.05,
help=
"Multiplicative factor for noise added to polynomial computation for irrelevant features"
)
parser.add_argument("-noise_type",
choices=[
constants.ADDITIVE_GAUSSIAN,
constants.EPSILON_IRRELEVANT, constants.NO_NOISE
],
default=constants.EPSILON_IRRELEVANT)
parser.add_argument(
"-hierarchy_type",
help="Choice of hierarchy to generate",
default=constants.CLUSTER_FROM_DATA,
choices=[constants.CLUSTER_FROM_DATA, constants.RANDOM])
parser.add_argument(
"-clustering_instance_count",
type=int,
help="If provided, uses this number of instances to "
"cluster the data to generate a hierarchy, allowing the hierarchy to remain same across multiple "
"sets of instances",
default=0)
parser.add_argument("-num_interactions",
type=int,
default=0,
help="number of interaction pairs in model")
parser.add_argument(
"-exclude_interaction_only_features",
help="exclude interaction-only features in model"
" in addition to linear + interaction features (default included)",
action="store_false",
dest="include_interaction_only_features")
parser.set_defaults(include_interaction_only_features=True)
parser.add_argument(
"-contiguous_node_names",
action="store_true",
help="enable to change node names in hierarchy "
"to be contiguous for better visualization (but creating mismatch between node names and features indices)"
)
# Arguments used to qualify output directory, then passed to mihifepe.master
parser.add_argument("-perturbation",
default=constants.SHUFFLING,
choices=[constants.ZEROING, constants.SHUFFLING])
parser.add_argument("-num_shuffling_trials",
type=int,
default=100,
help="Number of shuffling trials to average over, "
"when shuffling perturbations are selected")
parser.add_argument("-analyze_interactions",
help="enable analyzing interactions",
action="store_true")
args, pass_args = parser.parse_known_args()
pass_args = " ".join(pass_args)
if not args.output_dir:
args.output_dir = (
"sim_outputs_inst_%d_feat_%d_noise_%.3f_relfraction_%.3f_pert_%s_shufftrials_%d"
% (args.num_instances, args.num_features, args.noise_multiplier,
args.fraction_relevant_features, args.perturbation,
args.num_shuffling_trials))
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
args.rng = np.random.RandomState(args.seed)
args.logger = utils.get_logger(__name__,
"%s/simulation.log" % args.output_dir)
pipeline(args, pass_args)