tfr.keras.losses.ApproxNDCGLoss

Computes approximate NDCG loss between y_true and y_pred.

tfr.keras.losses.ApproxNDCGLoss(
    reduction: tf.losses.Reduction = tf.losses.Reduction.AUTO,
    name: Optional[str] = None,
    lambda_weight: Optional[losses_impl._LambdaWeight] = None,
    temperature: float = 0.1,
    ragged: bool = False
)

Implementation of ApproxNDCG loss (Qin et al, 2008; Bruch et al, 2019). This loss is an approximation for tfr.keras.metrics.NDCGMetric. It replaces the non-differentiable ranking function in NDCG with a differentiable approximation based on the logistic function.

For each list of scores s in y_pred and list of labels y in y_true:

loss = sum_i (2^y_i - 1) / log_2(1 + approxrank(s_i))
approxrank(s_i) = 1 + sum_j (1 / (1 + exp(-(s_j - s_i) / temperature)))

Standalone usage:

y_true = [[1., 0.]]
y_pred = [[0.6, 0.8]]
loss = tfr.keras.losses.ApproxNDCGLoss()
loss(y_true, y_pred).numpy()
-0.655107

# Using ragged tensors
y_true = tf.ragged.constant([[1., 0.], [0., 1., 0.]])
y_pred = tf.ragged.constant([[0.6, 0.8], [0.5, 0.8, 0.4]])
loss = tfr.keras.losses.ApproxNDCGLoss(ragged=True)
loss(y_true, y_pred).numpy()
-0.80536866

Usage with the compile() API:

model.compile(optimizer='sgd', loss=tfr.keras.losses.ApproxNDCGLoss())

Definition:

\[ \mathcal{L}(\{y\}, \{s\}) = - \frac{1}{\text{DCG}(y, y)} \sum_{i} \frac{2^{y_i} - 1}{\log_2(1 + \text{rank}_i)} \]

where:

\[ \text{rank}_i = 1 + \sum_{j \neq i} \frac{1}{1 + \exp\left(\frac{-(s_j - s_i)}{\text{temperature} }\right)} \]

References
A General Approximation Framework for Direct Optimization of Information Retrieval Measures, Qin et al, 2008 Revisiting Approximate Metric Optimization in the Age of Deep Neural Networks, Bruch et al, 2019

References

Args
`reduction`	Type of `tf.keras.losses.Reduction` to apply to loss. Default value is `AUTO`. `AUTO` indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to `SUM_OVER_BATCH_SIZE`. When used under a `tf.distribute.Strategy`, except via `Model.compile()` and `Model.fit()`, using `AUTO` or `SUM_OVER_BATCH_SIZE` will raise an error. Please see this custom training tutorial for more details.
`name`	Optional name for the instance.

Args

reduction Type of tf.keras.losses.Reduction to apply to loss. Default value is AUTO. AUTO indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to SUM_OVER_BATCH_SIZE. When used under a tf.distribute.Strategy, except via Model.compile() and Model.fit(), using AUTO or SUM_OVER_BATCH_SIZE will raise an error. Please see this custom training tutorial for more details.

name Optional name for the instance.

Methods

`from_config`

View source

@classmethod
from_config(
    config, custom_objects=None
)

Instantiates a Loss from its config (output of get_config()).

Args
`config`	Output of `get_config()`.

Returns
A `Loss` instance.

`get_config`

View source

get_config() -> Dict[str, Any]

Returns the config dictionary for a Loss instance.

`call`

View source

__call__(
    y_true: tfr.keras.model.TensorLike,
    y_pred: tfr.keras.model.TensorLike,
    sample_weight: Optional[utils.TensorLike] = None
) -> tf.Tensor

See tf.keras.losses.Loss.