Method _local_connection2d_update

bindsnet/learning/learning.py:287–349

Post-pre learning rule for ``LocalConnection2D`` subclass of ``AbstractConnection`` class.

(self, **kwargs)

        super().update()

    def _local_connection2d_update(self, **kwargs) -> None:
        # language=rst
        """
        Post-pre learning rule for ``LocalConnection2D`` subclass of
        ``AbstractConnection`` class.
        """
        # Get LC layer parameters.
        stride = self.connection.stride
        batch_size = self.source.batch_size
        kernel_height = self.connection.kernel_size[0]
        kernel_width = self.connection.kernel_size[1]
        in_channels = self.connection.source.shape[0]
        out_channels = self.connection.n_filters
        height_out = self.connection.conv_size[0]
        width_out = self.connection.conv_size[1]

        target_x = self.target.x.reshape(
            batch_size, out_channels * height_out * width_out, 1
        )
        target_x = target_x * torch.eye(out_channels * height_out * width_out).to(
            self.connection.w.device
        )
        source_s = (
            self.source.s.type(torch.float)
            .unfold(-2, kernel_height, stride[0])
            .unfold(-2, kernel_width, stride[1])
            .reshape(
                batch_size,
                height_out * width_out,
                in_channels * kernel_height * kernel_width,
            )
            .repeat(1, out_channels, 1)
            .to(self.connection.w.device)
        )

        target_s = self.target.s.type(torch.float).reshape(
            batch_size, out_channels * height_out * width_out, 1
        )
        target_s = target_s * torch.eye(out_channels * height_out * width_out).to(
            self.connection.w.device
        )
        source_x = (
            self.source.x.unfold(-2, kernel_height, stride[0])
            .unfold(-2, kernel_width, stride[1])
            .reshape(
                batch_size,
                height_out * width_out,
                in_channels * kernel_height * kernel_width,
            )
            .repeat(1, out_channels, 1)
            .to(self.connection.w.device)
        )

        # Pre-synaptic update.
        if self.nu[0].any():
            pre = self.reduction(torch.bmm(target_x, source_s), dim=0)
            self.connection.w -= self.nu[0] * pre.view(self.connection.w.size())
        # Post-synaptic update.