• namespace: Rindow\NeuralNetworks\Layer
• classname: Concatenate

Layer that concatenates a list of inputs.

Combines the listed NDArrays on the specified axis and outputs.

Concatenate implements the operation:

• output := concat([input1, input2….], axis=-1)

Methods

constructor

$builer->Concatenate(
    int $axis=-1,
    array $input_shapes=null,
    string $name=null,
)

You can create a Concatenate layer instances with the Layer Builder.

Options

• axis: Axis to join.
• input_shape: Tell the first layer the shape of the input data. In input_shape, the batch dimension is not included.

Input shape

All shapes except the concatenate axis must be the same.

Output shape

The concatenate axis is the sum of each array, and the other shapes are the same as the input.

$concat = $builder->layers()->Concatenate();
....
$a = $mo->ones([4,3,2]);
$b = $mo->ones([4,3,3]);
$c = $mo->ones([4,3,4]);
....
$outputs = $concat->forward([$a,$b,$c],true);
# $outputs->shape() : [4,3,9]

Example of usage

class Foo extends AbstractModel
{
    public function __construct($backend,$builder)
    {
        ...
        $this->concat = $builder->layers()->Concatenate();
        ....
    }

    protected function call(.....) : NDArray
    {
        ...
        $outputs = $this->concat->forward([$input1, $input2],$training);
        ...
    }
}