ef45b5da8d
* Doc Migration from Gitlab (#1289) * doc migration * fix * Update FakeQuantize_1.md * Update performance_benchmarks.md * Updates graphs for FPGA * Update performance_benchmarks.md * Change DL Workbench structure (#1) * Changed DL Workbench structure * Fixed tags * fixes * Update ie_docs.xml * Update performance_benchmarks_faq.md * Fixes in DL Workbench layout * Fixes for CVS-31290 * [DL Workbench] Minor correction * Fix for CVS-30955 * Added nGraph deprecation notice as requested by Zoe * fix broken links in api doxy layouts * CVS-31131 fixes * Additional fixes * Fixed POT TOC * Update PAC_Configure.md PAC DCP 1.2.1 install guide. * Update inference_engine_intro.md * fix broken link * Update opset.md * fix * added opset4 to layout * added new opsets to layout, set labels for them * Update VisionAcceleratorFPGA_Configure.md Updated from 2020.3 to 2020.4 Co-authored-by: domi2000 <domi2000@users.noreply.github.com>
72 lines
2.9 KiB
Markdown
72 lines
2.9 KiB
Markdown
## NormalizeL2 <a name="NormalizeL2"></a> {#openvino_docs_ops_normalization_NormalizeL2_1}
|
|
|
|
**Versioned name**: *NormalizeL2-1*
|
|
|
|
**Category**: *Normalization*
|
|
|
|
**Short description**: *NormalizeL2* operation performs L2 normalization of the 1st input tensor in slices specified by the 2nd input.
|
|
|
|
**Attributes**
|
|
|
|
* *eps*
|
|
|
|
* **Description**: *eps* is the number to be added/maximized to/with the variance to avoid division by zero when normalizing the value. For example, *eps* equal to 0.001 means that 0.001 is used if all the values in normalization are equal to zero.
|
|
* **Range of values**: a positive floating-point number
|
|
* **Type**: `float`
|
|
* **Default value**: None
|
|
* **Required**: *yes*
|
|
|
|
* *eps_mode*
|
|
|
|
* **Description**: Specifies how *eps* is combined with L2 value calculated before division.
|
|
* **Range of values**: `add`, `max`
|
|
* **Type**: `string`
|
|
* **Default value**: None
|
|
* **Required**: *yes*
|
|
|
|
**Inputs**
|
|
|
|
* **1**: `data` - input tensor to be normalized. Type of elements is any floating point type. Required.
|
|
|
|
* **2**: `axes` - scalar or 1D tensor with axis indices for the `data` input along which L2 reduction is calculated. Required.
|
|
|
|
**Outputs**
|
|
|
|
* **1**: Tensor of the same shape and type as the `data` input and normalized slices defined by `axes` input.
|
|
|
|
**Detailed Description**
|
|
|
|
Each element in the output is the result of division of corresponding element from the `data` input tensor by the result of L2 reduction along dimensions specified by the `axes` input:
|
|
|
|
output[i0, i1, ..., iN] = x[i0, i1, ..., iN] / sqrt(eps_mode(sum[j0,..., jN](x[j0, ..., jN]**2), eps))
|
|
|
|
Where indices `i0, ..., iN` run through all valid indices for the 1st input and summation `sum[j0, ..., jN]` have `jk = ik` for those dimensions `k` that are not in the set of indices specified by the `axes` input of the operation. One of the corner cases is when `axes` is an empty list, then we divide each input element by itself resulting value 1 for all non-zero elements. Another corner case is where `axes` input contains all dimensions from `data` tensor, which means that a single L2 reduction value is calculated for entire input tensor and each input element is divided by that value.
|
|
|
|
`eps_mode` selects how the reduction value and `eps` are combined. It can be `max` or `add` depending on `eps_mode` attribute value.
|
|
|
|
**Example**
|
|
|
|
```xml
|
|
<layer id="1" type="NormalizeL2" ...>
|
|
<data eps="1e-8" eps_mode="add"/>
|
|
<input>
|
|
<port id="0">
|
|
<dim>6</dim>
|
|
<dim>12</dim>
|
|
<dim>10</dim>
|
|
<dim>24</dim>
|
|
</port>
|
|
<port id="1">
|
|
<dim>2</dim> <!-- value is [2, 3] that means independent normalization in each channel -->
|
|
</port>
|
|
</input>
|
|
<output>
|
|
<port id="2">
|
|
<dim>6</dim>
|
|
<dim>12</dim>
|
|
<dim>10</dim>
|
|
<dim>24</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
``` |