Input shape specification in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does 'input shape' mean in a TensorFlow model?

Input shape is the size and structure of the data that the model expects to receive. It tells the model how many features or dimensions each input example has.

beginner

How do you specify the input shape in a TensorFlow Keras Sequential model?

You specify the input shape in the first layer using the 'input_shape' argument, for example: Dense(10, input_shape=(5,)) means each input has 5 features.

intermediate

Why is the batch size not included in the input shape specification?

Batch size is flexible and can change during training or prediction. Input shape only describes the shape of one example, excluding the batch size.

beginner

What is the input shape for a grayscale image of size 28x28 pixels?

The input shape is (28, 28, 1) where 28x28 is the image size and 1 is the number of color channels (grayscale).

intermediate

How does input shape differ for sequence data like text or time series?

For sequence data, input shape usually includes the sequence length and the number of features per step, for example (100, 1) for 100 time steps with 1 feature each.

In TensorFlow Keras, how do you specify the input shape for a model expecting 10 features per example?

A input_shape=(10,)

B input_shape=10

C input_shape=[10]

D input_shape=(1,10)

The input shape should be a tuple describing the shape of one example, so (10,) means 10 features.

Which part of the input shape is flexible and usually not specified in the model?

A Number of features

B Number of channels

C Sequence length

D Batch size

Batch size can vary during training or prediction, so it is not fixed in the input shape.

What input shape would you use for a color image of size 64x64 pixels?

A (3, 64, 64)

B (64, 64)

C (64, 64, 3)

D (64, 3)

Color images usually have 3 channels (red, green, blue), so the shape is (height, width, channels).

For a sequence of 50 time steps with 2 features each, what is the input shape?

A (2,)

B (50, 2)

C (50,)

D (2, 50)

The input shape for sequence data is (sequence length, features per step).

If you forget to specify input shape in the first layer of a Sequential model, what happens?

A Model will automatically infer input shape from data

B Model will raise an error when building

C Model will use default input shape (1,)

D Model will train but with random input shape

Keras can infer input shape from the first batch of data if not specified explicitly.

Explain how to specify input shape for different types of data in TensorFlow models.

Think about images, tabular data, and sequences.

Describe why batch size is not part of the input shape and how it affects model training.

Consider how data is fed in groups during training.