def activate(self): " Updates neuron's state for a single time-step. " if(len(self._synapses) > 0): self.__update_state() return nn.sigmoid(self.__state + self._bias, self._response, self._activation_type) else: return self._output # in case it's a sensor
def __init__(self, neurontype, id = None, bias = 0.0, response = 1.0, activation_type = 'exp', tau = 1.0): super(CTNeuron, self).__init__(neurontype, id, bias, response, activation_type) # decay rate self.__tau = tau # needs to set the initial state (initial condition for the ODE) self.__state = 0.1 #TODO: Verify what's the "best" initial state # fist output self._output = nn.sigmoid(self.__state + self._bias, self._response, self._activation_type) # integration step self.__dt = 0.05 # depending on the tau constant, the integration step must
def __init__(self, neurontype, id=None, bias=0.0, response=1.0, activation_type='exp', tau=1.0): super(CTNeuron, self).__init__(neurontype, id, bias, response, activation_type) # decay rate self.__tau = tau # needs to set the initial state (initial condition for the ODE) self.__state = 0.1 #TODO: Verify what's the "best" initial state # fist output self._output = nn.sigmoid(self.__state + self._bias, self._response, self._activation_type) # integration step self.__dt = 0.05 # depending on the tau constant, the integration step must
def set_init_state(self, state): self.__state = state self._output = nn.sigmoid(self.__state + self._bias, self._response, self._activation_type)
def activate(self): """ Updates neuron's state for a single time-step. . """ assert self._type is not 'INPUT' self.__update_state() return nn.sigmoid(self.__state + self._bias, self._response, self._activation_type)