def nn_fitness(synapses, graph=False):
"""Returns the fitness of a set of neural network synapses. Shows a matrix
graph if graph is true. This fitness tries to reach a checkerboard pattern
"""
# Determine fitness by running the synapse values on a neural network
neuron_values = matrix_create(10, 10)
for i in range(10):
neuron_values[0, i] = 0.5
for i in range(9):
update(neuron_values, synapses, i)
dist_to_goal = dist(NEURON_GOAL, neuron_values[9, :].A.reshape(-1).tolist())
# Since neuron values are in [0, 1], max distance is sqrt(10)
norm_dist = dist_to_goal / math.sqrt(10)
if graph:
pp.imshow(neuron_values, cmap=pp.cm.gray, aspect='auto',
interpolation='nearest')
pp.show()
return 1.0 - norm_dist
def nn_fitness2(synapses, graph=False):
"""Returns the fitness of a set of neural network synapses. Shows a matrix
graph if graph is true
"""
# Determine fitness by running the synapse values on a neural network
neuron_values = matrix_create(10, 10)
for i in range(10):
neuron_values[0, i] = 0.5
for i in range(9):
update(neuron_values, synapses, i)
diff = 0.0
for i in range(0,9):
for j in range(0,9):
diff += abs(neuron_values[i, j] - neuron_values[i, j + 1])
diff += abs(neuron_values[i + 1, j] - neuron_values[i, j])
if graph:
pp.imshow(neuron_values, cmap=pp.cm.gray, aspect='auto',
interpolation='nearest')
pp.show()
return diff / 162.0
def nn_fitness(synapses, graph=False):
"""Returns the fitness of a set of neural network synapses. Shows a matrix
graph if graph is true. This fitness tries to reach a checkerboard pattern
"""
# Determine fitness by running the synapse values on a neural network
neuron_values = matrix_create(10, 10)
for i in range(10):
neuron_values[0, i] = 0.5
for i in range(9):
update(neuron_values, synapses, i)
dist_to_goal = dist(NEURON_GOAL,
neuron_values[9, :].A.reshape(-1).tolist())
# Since neuron values are in [0, 1], max distance is sqrt(10)
norm_dist = dist_to_goal / math.sqrt(10)
if graph:
pp.imshow(neuron_values,
cmap=pp.cm.gray,
aspect='auto',
interpolation='nearest')
pp.show()
return 1.0 - norm_dist
def nn_fitness2(synapses, graph=False):
"""Returns the fitness of a set of neural network synapses. Shows a matrix
graph if graph is true
"""
# Determine fitness by running the synapse values on a neural network
neuron_values = matrix_create(10, 10)
for i in range(10):
neuron_values[0, i] = 0.5
for i in range(9):
update(neuron_values, synapses, i)
diff = 0.0
for i in range(0, 9):
for j in range(0, 9):
diff += abs(neuron_values[i, j] - neuron_values[i, j + 1])
diff += abs(neuron_values[i + 1, j] - neuron_values[i, j])
if graph:
pp.imshow(neuron_values,
cmap=pp.cm.gray,
aspect='auto',
interpolation='nearest')
pp.show()
return diff / 162.0