def output_from(self, input_states: iter) -> tuple:
"""Return output states responding to given input neuron states.
states: iterable of booleans, giving the state of input neurons.
return: iterable of booleans, giving the state of output neurons.
"""
# Atoms creation:
# - define the neural network
# - add an atom up/1 or down/1 foreach input neuron according to its state
input_atoms = self.neural_network + ''.join(
'up(' + str(idn) + ').'
# position in list gives the input neuron id
for idn, is_up in enumerate(input_states, start=MINIMAL_NEURON_ID)
if is_up)
LOGGER.debug('INPUT ATOMS: "' + input_atoms + '"')
# ASP solver call
model = solving.model_from(input_atoms, FILE_ASP_RUNNING)
LOGGER.debug('OUTPUT ATOMS: ' + str(model))
up_outputs = set(
int(atom.split('(')[1].strip(')')) for atom in model
if atom.startswith('up'))
min_output_neuron_id = self.min_output_neuron_id
ret = tuple(
idx in up_outputs
for idx in range(min_output_neuron_id, self.maximal_neuron_id + 1))
LOGGER.debug('OUTPUT STATES: ' + str(ret))
return ret
def output_from(self, input_states:iter) -> tuple:
"""Return output states responding to given input neuron states.
states: iterable of booleans, giving the state of input neurons.
return: iterable of booleans, giving the state of output neurons.
"""
# Atoms creation:
# - define the neural network
# - add an atom up/1 or down/1 foreach input neuron according to its state
input_atoms = self.neural_network + ''.join(
'up(' + str(idn) + ').'
# position in list gives the input neuron id
for idn, is_up in enumerate(input_states, start=MINIMAL_NEURON_ID)
if is_up
)
LOGGER.debug('INPUT ATOMS: "' + input_atoms + '"')
# ASP solver call
model = solving.model_from(input_atoms, FILE_ASP_RUNNING)
LOGGER.debug('OUTPUT ATOMS: ' + str(model))
up_outputs = set(int(atom.split('(')[1].strip(')'))
for atom in model if atom.startswith('up'))
min_output_neuron_id = self.min_output_neuron_id
ret = tuple(idx in up_outputs for idx in range(min_output_neuron_id,
self.maximal_neuron_id+1))
LOGGER.debug('OUTPUT STATES: ' + str(ret))
return ret
def cleaned(network_atoms: str) -> str:
"""Return a cleaned version of given neural network to remove
useless neuron, give an orientation to edges,..."""
model = solving.model_from(network_atoms, FILE_ASP_CLEANING)
assert model is not None, 'cleaning network lead to non existing model'
return '.'.join(model) + ('.' if len(model) else '')
def _run(self, atoms):
return model_from(atoms, FILE_ASP_RUNNING)
def _clean(self, atoms):
return model_from(atoms, FILE_ASP_CLEANING)
def cleaned(network_atoms:str) ->str:
"""Return a cleaned version of given neural network to remove
useless neuron, give an orientation to edges,..."""
model = solving.model_from(network_atoms, FILE_ASP_CLEANING)
assert model is not None, 'cleaning network lead to non existing model'
return '.'.join(model) + ('.' if len(model) else '')
