def learn_synthetic(input_directory, output_directory, runs, sample_size, processes, time_out, learn_options: LearnOptions):
commands = []
db = get_synthetic_db(input_directory)
for name in db.getall():
entry = db.get(name)
matching_samples = []
for sample in entry["samples"]:
if sample["sample_size"] == sample_size and len(matching_samples) < runs:
matching_samples.append(sample)
if len(matching_samples) != runs:
raise RuntimeError("Insufficient samples available, prepare more samples first")
for sample in matching_samples:
detail_learn_options = learn_options.copy()
detail_learn_options.domain = os.path.join(input_directory, "{}.density".format(name))
detail_learn_options.data = os.path.join(input_directory, sample["samples_file"])
detail_learn_options.labels = os.path.join(input_directory, sample["labels_file"])
export_file = "{}{sep}{}.{}.{}.result" \
.format( output_directory, name, sample_size, sample["seed"], sep=os.path.sep)
log_file = "{}{sep}{}.{}.{}.log" \
.format(output_directory, name, sample_size, sample["seed"], sep=os.path.sep)
if not os.path.exists(os.path.dirname(export_file)):
os.makedirs(os.path.dirname(export_file))
commands.append("incal-track {} --export {} --log {}"
.format(detail_learn_options.print_arguments(), export_file, log_file))
run_commands(commands, processes, time_out)
def learn_benchmark(runs, sample_size, processes, time_out, learn_options: LearnOptions):
# def filter1(entry):
# return "real_variables_count" in entry and entry["real_variables_count"] + entry["bool_variables_count"] <= 10
#
# count = 0
# boolean = 0
# for name, entry, density_filename in select_benchmark_files(filter1):
# if entry["bool_variables_count"] > 0:
# boolean += 1
# count += 1
#
# print("{} / {}".format(boolean, count))
#
# count = 0
# boolean = 0
# for name, entry, density_filename in select_benchmark_files(benchmark_filter):
# if entry["bool_variables_count"] > 0:
# boolean += 1
# count += 1
#
# print("{} / {}".format(boolean, count))
def learn_filter(_e):
return benchmark_filter(_e) and "samples" in _e
count = 0
problems_to_learn = []
for name, entry, density_filename in select_benchmark_files(learn_filter):
if len(entry["bounds"]) > 0:
best_ratio = min(rel_ratio(t[1]) for t in entry["bounds"])
if best_ratio <= 0.3:
qualifying = [t for t in entry["bounds"] if rel_ratio(t[1]) <= 0.3 and abs(rel_ratio(t[1]) - best_ratio) <= best_ratio / 5]
selected = sorted(qualifying, key=lambda x: get_bound_volume(x[0]))[0]
print(name, "\n", rel_ratio(selected[1]), best_ratio, selected[0], entry["bool_variables_count"])
count += 1
selected_samples = [s for s in entry["samples"]
if s["bounds"] == selected[0] and s["sample_size"] >= sample_size]
if len(selected_samples) < runs:
raise RuntimeError("Insufficient number of data set available ({} of {})"
.format(len(selected_samples), runs))
elif len(selected_samples) > runs:
selected_samples = selected_samples[:runs]
for selected_sample in selected_samples:
problems_to_learn.append((name, density_filename, selected_sample))
commands = []
for name, density_filename, selected_sample in problems_to_learn:
detail_learn_options = learn_options.copy()
detail_learn_options.domain = density_filename
detail_learn_options.data = selected_sample["samples_filename"]
detail_learn_options.labels = selected_sample["labels_filename"]
export_file = "{}{sep}{}.{}.{}.result".format(get_benchmark_results_dir(), name, selected_sample["sample_size"],
selected_sample["seed"], sep=os.path.sep)
log_file = "{}{sep}{}.{}.{}.log".format(get_benchmark_results_dir(), name, selected_sample["sample_size"],
selected_sample["seed"], sep=os.path.sep)
if not os.path.exists(os.path.dirname(export_file)):
os.makedirs(os.path.dirname(export_file))
commands.append("incal-track {} --export {} --log {}"
.format(detail_learn_options.print_arguments(), export_file, log_file))
run_commands(commands, processes, time_out)