def final_report():
alloc_gen, alloc_conf = importer.restore_genome("best_allocation_genome",
exp_dir)
alloc_ann = neat.nn.FeedForwardNetwork.create(alloc_gen, alloc_conf)
seq_gen, seq_conf = importer.restore_genome("best_sequencing_genome",
exp_dir)
seq_ann = neat.nn.FeedForwardNetwork.create(seq_gen, seq_conf)
print(
"\n### FINAL REPORT ###",
"\n\nBest allocation genome:\n{!s}".format(alloc_gen),
"\n\nBest sequencing genome:\n{!s}".format(seq_gen),
)
for dataset in all_datasets.items():
model = run_simulation(dataset[1], alloc_ann, seq_ann)
print(
"\n--- Performance on {} ------------------------".format(
dataset[0]),
"\nMakespan: ",
model.makespan,
"\nTotal tardiness: ",
model.total_tardiness,
"\nMajor setups: ",
model.num_major_setups,
"\nJobs processed: ",
model.jobs_processed,
"\n-----------------------------------------------------",
)
print(
"\nComputational time: ",
timeit.default_timer() - start,
"\nApproximal number of simulation runs (allocation | sequencing | overall): {} | {} | {}\n"
.format(num_runs_alloc, num_runs_seq, num_runs_alloc + num_runs_seq),
)
# visualize.draw_net(alloc_conf, alloc_gen, False, filename="alloc_net", objective=ALLOCATION_OBJECTIVE) # works only of graphviz binaries are installed
visualize.plot_stats(stats_alloc,
ylog=False,
view=False,
filename="avg_fitness_alloc.svg")
visualize.plot_species(stats_alloc,
view=False,
filename="speciation_alloc.svg")
# visualize.draw_net(seq_conf, seq_gen, False, filename="seq_net", objective=SEQUENCING_OBJECTIVE) # works only of graphviz binaries are installed
visualize.plot_stats(stats_seq,
ylog=False,
view=False,
filename="avg_fitness_seq.svg")
visualize.plot_species(stats_seq,
view=False,
filename="speciation_seq.svg")
def eval_fitness_seq(genome, config):
# Parse ANN
neat_ann = neat.nn.FeedForwardNetwork.create(genome, config)
# If possible: Parse allocation ANN from best genome of last session
try:
gen, conf = importer.restore_genome("best_allocation_genome", exp_dir)
alloc_ann = neat.nn.FeedForwardNetwork.create(gen, conf)
except:
alloc_ann = None
# Initialize fitness
fitness_overall = 0
# Evaluate NEAT genomes on datasets for training
for dataset in train_datasets.values():
model = run_simulation(dataset, alloc_ann, neat_ann)
fitness_overall += FITNESS_FUNCTION(model)
return fitness_overall
SMD_ALLOCATION_FUNCTION = alloc.smd_allocation_min_workload
if __name__ == "__main__":
print(
"\n-------------------------------------------------------------------------------------------------------------",
"\nSMD_ALLOCATION_FUNCTION automatically set to {}, because ANN_FOR_ALLOCATION is not"
.format(SMD_ALLOCATION_FUNCTION.__name__),
"\ngiven. The user selected SMD_ALLOCATION_FUNCTION is only valid if ANN_FOR_SEQUENCING is not None",
"\nor if the OBJECTIVE is '{}'".format(required_mode),
"\n-------------------------------------------------------------------------------------------------------------",
)
# pylint: enable=E1135
# Create ANNs
if ANN_FOR_SEQUENCING is not None:
gen, conf = importer.restore_genome(ANN_FOR_SEQUENCING, this_dir)
sequencing_ann = neat.nn.FeedForwardNetwork.create(gen, conf)
else:
sequencing_ann = None
if ANN_FOR_ALLOCATION is not None:
gen, conf = importer.restore_genome(ANN_FOR_ALLOCATION, this_dir)
allocation_ann = neat.nn.FeedForwardNetwork.create(gen, conf)
else:
allocation_ann = None
if __name__ == "__main__":
# Get config file
config_file = "config_{}".format(OBJECTIVE)
config_path = os.path.join(this_dir, config_file)
# Get current datetime