Versioned name: GRN-1

Category: Normalization

Short description: GRN is the Global Response Normalization with L2 norm (across channels only).

Detailed description:

GRN computes the L2 norm across channels for input tensor with shape [N, C, ...]. GRN does the following with the input tensor:

\[output[i0, i1, ..., iN] = x[i0, i1, ..., iN] / sqrt(sum[j = 0..C-1](x[i0, j, ..., iN]**2) + bias)\]


  • bias

    • Description: bias is added to the sum of squares.

    • Range of values: a positive floating-point number

    • Type: float

    • Required: yes


  • 1: data - A tensor of type T and 2 <= rank <= 4. Required.


  • 1: The result of GRN function applied to data input tensor. Normalized tensor of the same type and shape as the data input.


  • T: arbitrary supported floating-point type.


<layer ... type="GRN">
    <data bias="1e-4"/>
        <port id="0">
        <port id="0" precision="f32">