The following code was used when compiling the model: model.compile(optimizer='Adam,loss='categorical.crossentropy',metrics=[tf.keras.metrics.accurac y]), currently using evaluate When the method…

Question

Options

• Aaccuracy
• Bcategorical_ 1oss
• Closs
• Dcategorical accuracy

• A
  85% (17)
• B
  10% (2)
• D
  5% (1)

Explanation

When model.evaluate() runs, Keras always outputs two categories of values: (1) the loss value, always labeled "loss" regardless of which loss function was chosen, and (2) each metric passed to metrics=[] using its display name - tf.keras.metrics.accuracy registers under the name "accuracy". This is why A (accuracy) and C (loss) are both correct.

Why the distractors fail:

• B (categorical_loss) - this name does not exist in Keras; the loss output is always labeled "loss", never prefixed with the loss function's name.
• D (categorical_accuracy) - this would only appear if you compiled with metrics=['categorical_accuracy'] or tf.keras.metrics.CategoricalAccuracy(); the generic tf.keras.metrics.accuracy outputs as "accuracy", not "categorical_accuracy".

Memory tip: Think of model.evaluate() output as two slots - slot 1 is always loss (the raw loss value, plain label), and slot 2+ are your metrics by their Keras display name. The loss function's full name (categorical_crossentropy) never appears in the output; only the word loss does.

No community discussion yet for this question.