tf.keras.ops.in_top_k

Checks if the targets are in the top-k predictions.

tf.keras.ops.in_top_k(
    targets, predictions, k
)

targets A tensor of true labels.
predictions A tensor of predicted labels.
k An integer representing the number of predictions to consider.

A boolean tensor of the same shape as targets, where each element indicates whether the corresponding target is in the top-k predictions.

Example:

targets = keras.ops.convert_to_tensor([2, 5, 3])
predictions = keras.ops.convert_to_tensor(
[[0.1, 0.4, 0.6, 0.9, 0.5],
 [0.1, 0.7, 0.9, 0.8, 0.3],
 [0.1, 0.6, 0.9, 0.9, 0.5]])
in_top_k(targets, predictions, k=3)
array([ True False  True], shape=(3,), dtype=bool)