def eval_genomes(genomes, config):
# Play game and get results
idx, genomes = zip(*genomes)
MyNeatFlappy = FlappyBirdApp(genomes, config)
MyNeatFlappy.play()
results = MyNeatFlappy.crash_info
# Calculate fitness and top score
top_score = 0
for result, genomes in results:
score = result['score']
distance = result['distance']
energy = result['energy']
fitness = score * 3000 + 0.2 * distance - energy * 1.5
if fitness == 0:
genomes.fitness = -1
else:
genomes.fitness = fitness
if top_score < score:
top_score = score
#print score
print('The top score was:', top_score)
def run_winner(n=1):
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
'flappy-config')
genomes = pickle.load(open('winner.pkl', 'rb'))
for i in range(0, n):
# Play game and get results
flappy_Bio = FlappyBirdApp([genomes], config)
flappy_Bio.play()
def eval_genomes(genomes, config):
idx, genomes = zip(*genomes)
flappy_Bio = FlappyBirdApp(genomes, config)
flappy_Bio.play()
results = flappy_Bio.crash_info
top_score = 0
for result, genomes in results:
score = result['score']
distance = result['distance']
energy = result['energy']
fitness = score * 3000 + 0.2 * distance - energy * 1.5
genomes.fitness = -1 if fitness == 0 else fitness
if top_score < score:
top_score = score
print('The top score was:', top_score)
def fitness(self, current_generation):
# Play game and get results
flappy_Bio = FlappyBirdApp(current_generation)
flappy_Bio.play()
results = flappy_Bio.crash_info
# Calculate fitness
self.fitness_values = []
self.network_index_dict = {}
for index, result in enumerate(results):
network = result['network']
self.network_index_dict[index] = network
distance = result['distance']
energy = result['energy']
fitness = distance - energy * 0.5
fitness = -1 if fitness == 0 else fitness
self.fitness_values.append((index, fitness))
def fitness(self):
"""
Fitness
-------
Each species in the pool is iterated over.
They are passed into the FlappyBird App,
and the game is played for all organisms in that species simultaneously.
Flappy generates "crash_info", from which we obtain the
1. organism,
2. its expended energy during play, and
3. its distance traveled during play
Using the organism's energy and distance, a fitness measure is generated and assigned to the organism
"""
# Iterate through the species in the list
for species in self.species:
# ================ PLAY FLAPPYBIRD ====================
flappy = FlappyBirdApp(species, self.generation)
flappy.play()
# =====================================================
for bird_results in flappy.crash_info:
# ====== Results per organism =======
organism = bird_results['network']
energy = bird_results['energy']
distance = bird_results['distance']
# ===================================
# ======== INTERFACE WITH VIS ========
print(organism)
# ===================================
# Determine the organism's fitness
# ============ FITNESS FUNCTION ================
fitness = distance - energy * 1.5
# ==============================================
# Assign the fitness to the organism
organism.fitness = fitness if fitness > 0 else -1.0
# =============== Pool's Max Fitness ===============
# First, check if max fitness dict is empty.
# If so, assign the first organism and its fitness
pool_max_organism = self.max_fitness['organism']
if not pool_max_organism['fitness']:
pool_max_organism['object'] = organism
pool_max_organism['fitness'] = organism.fitness
# Check if current organism has higher fitness than pool's all-time max
elif organism.fitness > pool_max_organism['fitness']:
pool_max_organism['object'] = organism
pool_max_organism['fitness'] = organism.fitness
# ==================================================
# Set the species intra rank and reorder the species organism list from top rank to lowest rank (0 to x, x being the lowest)
species.set_intra_species_rank()