Flatten and Dense layers in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does a Flatten layer do in a neural network?

A Flatten layer takes a multi-dimensional input (like an image) and turns it into a single long list of numbers. This helps connect image data to layers that expect flat input.

beginner

What is the purpose of a Dense layer in a neural network?

A Dense layer connects every input to every output neuron with weights. It helps the network learn patterns by combining inputs in different ways.

beginner

Why do we often use a Flatten layer before a Dense layer?

Because Dense layers expect flat input, we use Flatten to convert multi-dimensional data (like images) into a flat list so the Dense layer can process it.

beginner

In TensorFlow, how do you add a Flatten layer to a model?

You add it like this: tf.keras.layers.Flatten(). It will reshape the input to a flat vector automatically.

intermediate

What activation function is commonly used in Dense layers for classification?

The 'softmax' activation is often used in the last Dense layer for multi-class classification to output probabilities.

What shape does a Flatten layer output if the input shape is (28, 28, 1)?

A (28, 28, 1)

B (28, 28)

C (784,)

D (1, 784)

Flatten converts the input to a single dimension by multiplying all dimensions: 28*28*1 = 784.

Which layer type connects every input neuron to every output neuron?

A Dense

B Pooling

C Convolutional

D Flatten

Dense layers have full connections between inputs and outputs.

Why is Flatten needed before Dense layers when working with images?

A To convert multi-dimensional data into a flat vector

B To reduce the number of neurons

C To increase image size

D To apply activation functions

Dense layers require flat input, so Flatten reshapes the image data.

Which activation function is commonly used in the last Dense layer for multi-class classification?

A Sigmoid

B Softmax

C ReLU

D Tanh

Softmax outputs probabilities for each class, summing to 1.

In TensorFlow, how do you add a Dense layer with 10 neurons and ReLU activation?

A tf.keras.layers.Dense(activation='relu')

B tf.keras.layers.Flatten(10, activation='relu')

C tf.keras.layers.Dense(10)

D tf.keras.layers.Dense(10, activation='relu')

You specify the number of neurons and activation function inside Dense.

Explain in your own words why we use a Flatten layer before Dense layers in image classification models.

Think about how images are shaped and what Dense layers expect.

Describe the role of Dense layers in a neural network and how they learn from data.

Consider how neurons communicate and learn connections.