1be7c5996fa959572854044d3cc580a2abc6bfa7,snntoolbox/core/inisim.py,SpikeMaxPooling2D,call,#SpikeMaxPooling2D#Any#Any#,559

Before Change

    def call(self, x, mask=None):
        Layer functionality.

        inp = x[0]

        if settings["payloads"]:
            // Add payload from previous layer
            inp = add_payloads(self.inbound_nodes[0].inbound_layers[0], inp)

        if "binary" in settings["maxpool_type"]:
            self.impulse = super(SpikeMaxPooling2D, self).call(inp)
        elif settings["maxpool_type"] in ["avg_max", "fir_max", "exp_max"]:
            spikerate = self.inbound_nodes[0].inbound_layers[0].spikerate \
                if self.inbound_nodes[0].inbound_layers \
                else k.placeholder(shape=self.input_shape) // Test if needed.
            max_idx = pool_same_size(spikerate, self.pool_size,
                                     self.ignore_border, self.strides)
            self.impulse = super(SpikeMaxPooling2D, self).call(t.mul(inp,
                                                                     max_idx))
        else:
            print("Wrong max pooling type, "
                  "falling back on Average Pooling instead.")
            self.impulse = k.pool2d(inp, self.pool_size, self.strides,
                                    self.border_mode, pool_mode="avg")

        output_spikes = update_neurons(self, x)
        return t.cast(output_spikes, "float32")

    @property
    def class_name(self):
        Get class name.

After Change

            self.impulse = k.pool2d(inp, self.pool_size, self.strides,
                                    self.border_mode, pool_mode="avg")

        return update_neurons(self, x)

    def reset(self):
        Reset layer variables.

Instances