npair_loss¶

paddle.nn.functional. npair_loss ( anchor: Tensor, positive: Tensor, labels: Tensor, l2_reg: float = 0.002 ) → Tensor [source]

Npair loss requires paired data. Npair loss has two parts: the first part is L2 regularizer on the embedding vector; the second part is cross entropy loss which takes the similarity matrix of anchor and positive as logits.

For more information, please refer to: Improved Deep Metric Learning with Multi class N pair Loss Objective

Parameters

• anchor (Tensor) – embedding vector for the anchor image. shape=[batch_size, embedding_dims], the data type is float32 or float64.

• positive (Tensor) – embedding vector for the positive image. shape=[batch_size, embedding_dims], the data type is float32 or float64.

• labels (Tensor) – 1-D tensor. shape=[batch_size], the data type is float32 or float64 or int64.

• l2_reg (float, optional) – L2 regularization term on embedding vector, default: 0.002.

Returns

A 0-D Tensor representing the npair loss, the data type is the same as anchor, the shape is [].

Examples

>>> import paddle
>>> from typing import Literal
>>> paddle.seed(2023)
>>> dtype: Literal["float32"] = "float32"

>>> anchor = paddle.rand(shape=(18, 6), dtype=dtype)
>>> positive = paddle.rand(shape=(18, 6), dtype=dtype)
>>> labels = paddle.rand(shape=(18,), dtype=dtype)

>>> npair_loss = paddle.nn.functional.npair_loss(anchor, positive, labels, l2_reg = 0.002)
>>> print(npair_loss)
Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True,
        2.94269347)