Neural Network Deep Learning Architectures 1 β Questions and Answers
Question 1: What distinguishes a deep neural network from a shallow neural network?
- It uses GPU acceleration
- It has multiple hidden layers (Correct answer)
- It requires more training data
- It uses a different loss function
Correct answer: It has multiple hidden layers
A deep neural network contains multiple hidden layers that allow it to learn hierarchical representations of data.
Question 2: Which architecture is specifically designed for processing sequential data with long-range dependencies?
- Convolutional Neural Network
- Autoencoder
- Transformer (Correct answer)
- Restricted Boltzmann Machine
Correct answer: Transformer
The Transformer architecture uses self-attention mechanisms to handle long-range dependencies in sequential data without recurrence.
Question 3: What is the primary role of the encoder in an autoencoder network?
- To classify input data
- To compress input into a latent representation (Correct answer)
- To generate new samples
- To apply dropout regularization
Correct answer: To compress input into a latent representation
The encoder compresses the input data into a lower-dimensional latent space representation.
Question 4: In a Generative Adversarial Network (GAN), what is the role of the discriminator?
- To generate realistic samples
- To encode latent variables
- To distinguish real samples from generated ones (Correct answer)
- To apply weight regularization
Correct answer: To distinguish real samples from generated ones
The discriminator learns to classify whether a given sample is real (from the training set) or fake (generated by the generator).
Question 5: Which component of a Variational Autoencoder (VAE) enables backpropagation through the sampling step?
- The KL divergence term
- The reparameterization trick (Correct answer)
- The reconstruction loss
- The encoder bottleneck
Correct answer: The reparameterization trick
The reparameterization trick expresses the random sample as a deterministic function of the latent mean and variance, enabling gradients to flow through.
Question 6: What is residual learning in the context of deep neural networks?
- Learning from leftover training data
- Adding skip connections so layers learn residual mappings (Correct answer)
- Using dropout to reduce overfitting
- Training with smaller learning rates
Correct answer: Adding skip connections so layers learn residual mappings
Residual learning adds skip connections that allow layers to learn the residual (difference) rather than the full mapping, easing training of very deep networks.
Neural Network Deep Learning Architectures 1
What distinguishes a deep neural network from a shallow neural network?