class MapConnectionGene(BaseGene):
#Various parameters for defining a connection.
_gene_attributes = [
BoolAttribute('one_to_one'), #1-to-1 or 1-to-all scheme
BoolAttribute('is_gaussian'), #Gaussian or uniform distribution
FloatAttribute(
'weight'
), #Weigth is used as the mean of the normal distribution for 1-to-all
FloatAttribute('sigma'), #The standard deviation for the gaussian
BoolAttribute('enabled')
] #<- maybe remove this trait
def __init__(self, key):
assert isinstance(
key, tuple
), "DefaultConnectionGene key must be a tuple, not {!r}".format(key)
BaseGene.__init__(self, key)
#Define the distance between two genes
def distance(self, other, config):
d = abs(self.sigma -
other.sigma) + abs(self.weight - other.weight) + int(
self.one_to_one != other.one_to_one)
+int(self.is_gaussian != other.is_gaussian)
return d * config.compatibility_weight_coefficient
class GuidedMapConnectionGene(BaseGene):
#Various parameters for defining a connection.
_gene_attributes = [
BoolAttribute(
'one2one'
), #if true then the connection scheme is one to one, else one to all
BoolAttribute('extended'),
BoolAttribute('uniform'), #step or uniform
FloatAttribute(
'weight'
), #Weigth is used as the mean of the normal distribution for 1-to-all
BoolAttribute('enabled'),
BoolAttribute('is_mod')
]
def __init__(self, key):
assert isinstance(
key, tuple
), "DefaultConnectionGene key must be a tuple, not {!r}".format(key)
BaseGene.__init__(self, key)
#Define the distance between two genes
def distance(self, other, config):
d = abs(self.weight - other.weight) + int(self.one2one == other.one2one) \
+ int(self.uniform == other.uniform) + int(self.enabled == other.enabled) + int(self.is_mod == other.is_mod) \
+ int(self.extended == other.extended)
return d * config.compatibility_weight_coefficient
class ExtendedMapNodeGene(DefaultNodeGene):
_gene_attributes = [FloatAttribute('bias'), #The bias of the neuron
StringAttribute('activation', options='sigmoid'), # The activation function, tunable from the config
StringAttribute('aggregation', options='sum'), #The aggregation function
BoolAttribute('is_isolated'), #Map vs isolated neuron
BoolAttribute('is_switch')]
def distance(self, other, config):
#Heavier weights for activation and aggregation functions - force neat to put identical networks with different]
#functions on the output neuron to different species
d = abs(self.bias - other.bias) + 5 * int(self.activation == other.activation) + 5 * int(self.aggregation == other.aggregation) \
+ int(self.is_isolated == other.is_isolated) + int(self.is_switch - other.is_switch)
return d * config.compatibility_weight_coefficient
class SwitchConnectionGene(DefaultConnectionGene):
_gene_attributes = [
FloatAttribute('weight'),
BoolAttribute('is_mod'),
BoolAttribute('enabled')
] #The neat package complains if this attribute is not present.
def distance(self, other, config):
d = abs(self.weight - other.weight)
if self.enabled != other.enabled:
d += 1.0
if self.is_mod != other.is_mod:
d += 1
return d * config.compatibility_weight_coefficient
class FormConnectionGene(BaseGene):
__gene_attributes__ = [
MatrixAttribute('transform', default=np.identity(4)),
# BoolAttribute('copy'),
BoolAttribute('negate'),
BoolAttribute('enabled')
]
def distance(self, other, config):
d = np.linalg.norm(self.transform - other.transform)
# d += self.copy != other.copy
d += self.negate != other.negate
d += self.enabled != other.enabled
return d * config.compatibility_weight_coefficient
class GuidedMapNodeGene(DefaultNodeGene):
_gene_attributes = [
FloatAttribute('bias'), #The bias of the neuron
StringAttribute('activation', options='sigmoid'
), # The activation function, tunable from the config
StringAttribute('aggregation',
options='sum'), #The aggregation function
BoolAttribute('is_isolated'), #Map vs isolated neuron
BoolAttribute('is_switch')
]
def distance(self, other, config):
d = abs(self.bias - other.bias) + int(self.activation == other.activation) + int(self.aggregation == other.aggregation) \
+ int(self.is_isolated == other.is_isolated) + int(self.is_switch - other.is_switch)
return d * config.compatibility_weight_coefficient
class TransitionGene(BaseGene):
""" Class representing the gene of a transition in the state machine. """
_gene_attributes = [
ConditionsAttribute('conditions'),
BoolAttribute('enabled')
]
def __init__(self, key):
assert isinstance(
key,
tuple), "TransitionGene key must be an tuple, not {!r}".format(key)
BaseGene.__init__(self, key)
def distance(self, other, config):
assert isinstance(other, TransitionGene)
# TODO: add difference with maximal value for difference of condition being 1.
# TODO: Track conditions (possibly)
return config.compatibility_difference_coefficient * abs(
len(other.conditions) - len(self.conditions))
def copy(self):
transition = TransitionGene(self.key)
transition.conditions = copy.deepcopy(self.conditions)
transition.enabled = self.enabled
return transition
class DefaultConnectionGene(BaseGene):
__gene_attributes__ = [FloatAttribute('weight'), BoolAttribute('enabled')]
def distance(self, other, config):
d = abs(self.weight - other.weight)
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class SharedConnectionGene(DefaultConnectionGene):
_gene_attributes = [EmptyAttribute('weight'), BoolAttribute('enabled')]
def __init__(self, key):
BaseGene.__init__(self, key)
def get_weight(self):
return self.weight.weight
class CircuitConnectionGene(BaseGene):
__gene_attributes__ = [StringAttribute('component'),
FloatAttribute('value'),
BoolAttribute('enabled')]
def distance(self, other, config):
d = abs(self.value - other.value)
if self.component != other.component:
d += 1.0
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class DefaultConnectionGene(BaseGene):
__gene_attributes__ = [FloatAttribute('weight'), BoolAttribute('enabled')]
@classmethod
def parse_config(cls, config, param_dict):
return DefaultGeneConfig(cls.__gene_attributes__, param_dict)
def distance(self, other, config):
d = abs(self.weight - other.weight)
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class DefaultConnectionGene(BaseGene):
_gene_attributes = [FloatAttribute('weight'),
BoolAttribute('enabled')]
def __init__(self, key):
assert isinstance(key, tuple), "DefaultConnectionGene key must be a tuple, not {!r}".format(key)
BaseGene.__init__(self, key)
def distance(self, other, config):
d = abs(self.weight - other.weight)
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class CircuitConnectionGene(BaseGene):
__gene_attributes__ = [
StringAttribute('component'),
FloatAttribute('value'),
BoolAttribute('enabled')
]
@classmethod
def parse_config(cls, config, param_dict):
return DefaultGeneConfig(cls.__gene_attributes__, param_dict)
def distance(self, other, config):
d = abs(self.value - other.value)
if self.component != other.component:
d += 1.0
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class SwitchNodeGene(DefaultNodeGene):
_gene_attributes = [
FloatAttribute('bias'),
StringAttribute('activation', options='sigmoid'),
StringAttribute('aggregation', options='sum'),
BoolAttribute('is_switch')
]
def distance(self, other, config):
d = abs(self.bias + other.bias)
if self.activation != other.activation:
d += 1.0
if self.aggregation != other.aggregation:
d += 1.0
if self.is_switch != other.is_switch:
d = 3
return d * config.compatibility_weight_coefficient
class DefaultConnectionGene(BaseGene):
_gene_attributes: List[BaseAttribute] = [
FloatAttribute("weight"),
BoolAttribute("enabled"),
]
def __init__(self, key: Tuple[int, int]):
assert isinstance(
key, tuple
), "DefaultConnectionGene key must be a tuple, not {!r}".format(key)
BaseGene.__init__(self, key)
def distance(self, other: DefaultConnectionGene,
config: DefaultGenomeConfig) -> float:
d = abs(self.weight - other.weight)
if self.enabled != other.enabled:
d += 1.0
return d * config.compatibility_weight_coefficient
class MapNodeGene(DefaultNodeGene):
_gene_attributes = [
FloatAttribute('bias'), #The bias of the neuron
StringAttribute('activation', options='sigmoid'
), # The activation function, tunable from the config
StringAttribute('aggregation',
options='sum'), #The aggregation function
BoolAttribute('is_isolated')
] #Map vs isolated neuron
def distance(self, other, config):
d = 0
if self.activation != other.activation:
d += 1.0
if self.aggregation != other.aggregation:
d += 1.0
if self.is_isolated != other.is_isolated:
d += 1
return d * config.compatibility_weight_coefficient
