def crossval_small_dataset(shingles_type):
in_file = helpers.datasets[dataset]["files"][0]
graph_database = list(dataset_manager.read_chemical_compounts(in_file))
for window_size in window_size_range:
base_model = {"window_size": window_size}
# best_model = crossval.loo_crossval_sketch(graph_database, wl_iter_range, k_L_range, output_dir, cols_count=188, base_model=base_model, shingles_type=shingles_type, window_size=window_size, accumulate_wl_shingles=False)
best_model = crossval.loo_crossval_pnn(graph_database, wl_iter_range, p_range, output_dir, base_model=base_model, shingles_type=shingles_type, window_size=window_size, accumulate_wl_shingles=False)
# best_model = crossval.loo_crossval_threshold(graph_database, wl_iter_range, infl_point_range, output_dir, base_model=base_model, shingles_type=shingles_type, window_size=window_size, accumulate_wl_shingles=False)
print "Best model:", best_model
def loo_crossval(hypergraph, wl_iter_range, r_in_range, r_out_range, r_all_range, output_dir, k_L_range=None, infl_point_range=None, p_range=None):
best_model = sgm_crossval.model(-1, -1, base_model=model(-1, -1, -1))
for r_in in r_in_range:
for r_out in r_out_range:
for r_all in r_all_range:
base_model = model(r_in, r_out, r_all)
rballs_database, _ = similar_nodes_mining.extract_rballs_database(hypergraph, r_in=r_in, r_out=r_out, r_all=r_all, center_default_color=True)
rballs_database = [(r_id, list(graphs), t) for r_id, graphs, t in rballs_database] # execute generator
if k_L_range:
current_model = sgm_crossval.loo_crossval_sketch(rballs_database, wl_iter_range, k_L_range, output_dir, base_model=base_model)
elif infl_point_range:
current_model = sgm_crossval.loo_crossval_threshold(rballs_database, wl_iter_range, infl_point_range, output_dir, base_model=base_model)
else:
current_model = sgm_crossval.loo_crossval_pnn(rballs_database, wl_iter_range, p_range, output_dir, base_model=base_model)
if current_model["quality"] > best_model["quality"]:
best_model = current_model
models_file = open(output_dir + "models", "a")
models_file.write(str(best_model) + ",\n")
models_file.close()
return best_model