.. _op_ai_onnx_ml_Normalizer:

Normalizer
==========

- **Domain**: ``ai.onnx.ml``
- **Since version**: 1

Normalize the input.  There are three normalization modes, which have the corresponding formulas,
defined using element-wise infix operators '/' and '^' and tensor-wide functions 'max' and 'sum': 

Max: Y = X / max(X) 
L1:  Y = X / sum(X) 
L2:  Y = sqrt(X^2 / sum(X^2)} 
In all modes, if the divisor is zero, Y == X.

For batches, that is, [N,C] tensors, normalization is done along the C axis. In other words, each row
of the batch is normalized independently.

**Inputs**

- **X** (*T*): Data to be encoded, a tensor of shape [N,C] or [C]

**Outputs**

- **Y** (*tensor(float)*): Encoded output data

**Type Constraints**

- **T**: The input must be a tensor of a numeric type.
  Allowed types: tensor(double), tensor(float), tensor(int32), tensor(int64).

Examples
--------

**test_cc_normalizer_l1_int64**

.. code-block:: text

    Node:
      ai.onnx.ml.Normalizer(x) -> (y)
      Attributes:
        norm = "L1"

.. code-block:: text

    Inputs:
      x: shape=(4,), dtype=int64
        [ 1, -1,  2, -2]

    Outputs:
      y: shape=(4,), dtype=float32
        [ 0.16666667, -0.16666667,  0.33333334, -0.33333334]

**test_cc_normalizer_l2_float**

.. code-block:: text

    Node:
      ai.onnx.ml.Normalizer(x) -> (y)
      Attributes:
        norm = "L2"

.. code-block:: text

    Inputs:
      x: shape=(2, 3), dtype=float32
        [[1., 2., 2.],
         [0., 3., 4.]]

    Outputs:
      y: shape=(2, 3), dtype=float32
        [[0.33333334, 0.6666667 , 0.6666667 ],
         [0.        , 0.6       , 0.8       ]]

**test_cc_normalizer_max_double**

.. code-block:: text

    Node:
      ai.onnx.ml.Normalizer(x) -> (y)
      Attributes:
        norm = "MAX"

.. code-block:: text

    Inputs:
      x: shape=(2, 3), dtype=float64
        [[ 1., -3.,  2.],
         [ 0.,  0.,  0.]]

    Outputs:
      y: shape=(2, 3), dtype=float32
        [[ 0.5, -1.5,  1. ],
         [ 0. ,  0. ,  0. ]]