neuron.py 文件源码

def broadcast(self, input_, norm, classifier=None, aif_scale=1):
        """
        Used for high level broadcast, in which the input is processed in bulk instead of per-component, i.e., per byte.
        input_ of len 1-256 (for proper CM1K emulation, otherwise unlimited)
        norm: A DistNorm enum
        classifier: A Classifier enum
        aif_scale: Modify the aif when determining whether the fire. The aif can also be permanently scaled via
        CM1KEmulator.scale_all_aifs(), but this parameter enables the same behavior without altering the neuron.
        """
        # Called from CM1KEmulator.broadcast()
        log.trace("Neuron.broadcast()")

        # This shouldn't be necessary. This function should only be called on committed and the rtl neurons.
        if self.state == NeuronState.idle:
            log.error("Neuron.broadcast() called on idle neuron")
            return

        self.dist = 0  # NOTE: Not sure this is necessary. Also, undecided whether this should simply call reset_dist().
        if norm == cm1k.CM1KDistNorm.l1:
            for i, comp in enumerate(input_):
                self.dist += abs(comp - self.pattern[i])
        elif norm == cm1k.CM1KDistNorm.lsup:
            for i, comp in enumerate(input_):
                self.dist = max(abs(comp - self.pattern[i]), self.dist)
        elif norm == cm1k.CM1KDistNorm.euc:
            for i, comp in enumerate(input_):
                self.dist += (comp - self.pattern[i]) * (comp - self.pattern[i])
            self.dist = int(round(math.sqrt(self.dist)))
        log.log("Single neuron cat{} dist: {:>5} < {:>5} ?".format(self.cat, self.dist, self.aif))

        # TODO: Use the minimum and maximum AIFs of each neuron (i.e., of each context)
        aif = self.aif if aif_scale == 1 else min(max(int(round(self.aif * aif_scale)), 0), 0xFFFF)

        if (classifier == cm1k.CM1KClassifier.rbf and self.dist < aif) or classifier == cm1k.CM1KClassifier.knn:
            # The neuron has fired
            log.log("Fire with dist{} aif{} cat{}".format(self.dist, aif, self.cat))
            self.chip.store_firing_neuron(self)