Neural Network Deep Learning Architectures 2

Question 1

What problem does batch normalization primarily address in deep networks?

Accepted Answer

Internal covariate shift between layers

Answer

Batch normalization reduces internal covariate shift by normalizing layer inputs, stabilizing and accelerating training.

Question 2

Which architecture introduced the concept of inception modules combining multiple filter sizes in parallel?

Accepted Answer

GoogLeNet (Inception)

Answer

GoogLeNet introduced inception modules that apply 1×1, 3×3, and 5×5 convolutions in parallel to capture multi-scale features.

Question 3

What is the purpose of the bottleneck layer in an autoencoder?

Accepted Answer

To force the network to learn a compressed representation

Answer

The bottleneck layer constrains the network to learn a compressed, meaningful representation of the input by limiting the number of neurons.

Question 4

In a U-Net architecture, what is the purpose of the skip connections between encoder and decoder?

Accepted Answer

To preserve fine-grained spatial information during reconstruction

Answer

Skip connections in U-Net pass high-resolution feature maps from the encoder directly to the decoder, preserving spatial detail for precise segmentation.

Question 5

What distinguishes a capsule network from a standard CNN?

Accepted Answer

It encodes pose information as vectors rather than scalar activations

Answer

Capsule networks represent features as vectors (capsules) that encode both the presence and spatial relationships/pose of entities, unlike scalar activations in CNNs.

Neural Network Practice Test

Neural Network Practice Test

Neural Network Deep Learning Architectures 2