What’s the exact difference between the final prediction output of the preferred and non-preferred methods? Does the preferred method return a vector of z-values from 1 to N, and the non-preferred one returns a vector of probability a_1 to a_n?

Also, in the week 2 Softmax lab, I try to understand what the function: model.predict(X_train) returns. I know it has four columns because the target value of Y has four classes, but what do the rows stand for, and what factor influences the number of rows?

Thanks for helping!

What shape is your output from *.predict(input)* ?

What shape is your *input* ?

1 Like

I am also wondering what you mean by “preferred methods”. Are you talking about the `from_logits`

feature? Can you elaborate more?

Yes, what I mean by the preferred methods is the from_logits feature, the one makes the model more numerically stable.

That’s the problem, apparently, since it’s a tuple, I can’t call its shape.

When `"softmax"`

is specified in the output layer, we set `from_logits`

to `False`

.

When `"softmax"`

is specified in the output layer, the model outputs probabilities as you said.

The number of probabilities is equal to the number of classes.

When `"linear"`

is specified in the output layer, we set `from_logits`

to `True`

.

When `"linear"`

is specified in the output layer, the model outputs logits.

The number of logits is equal to the number of classes.

I think you are referring the logits as `z`

.

Number of samples to predict for.

Cheers,

Raymond

Neither the input nor the output of Keras model.predict() is typically a tuple for the exercises in these classes. See

###
`predict`

Args

`x`

Input samples. It could be:

- A Numpy array (or array-like), or a list of arrays (in case the model has multiple inputs).
- A TensorFlow tensor, or a list of tensors (in case the model has multiple inputs).
- A
`tf.data`

dataset.
- A generator or
`keras.utils.Sequence`

instance. A more detailed description of unpacking behavior for iterator types (Dataset, generator, Sequence) is given in the `Unpacking behavior for iterator-like inputs`

section of `Model.fit`

.

…

Returns

Numpy array(s) of predictions.

I had expected you would find that the input shape was something like (*m*,448,448,3) and the output shape (*m*,4). The clue, then, being the common first dimension which @rmwkwok points out is the number of inputs or samples.