Artificial Intelligence Neural Networks 3

Question 1

What is transfer learning in the context of neural networks?

Accepted Answer

Reusing a pre-trained model's weights as a starting point for a new task

Answer

Transfer learning leverages weights learned on a large dataset (e.g., ImageNet) and fine-tunes them for a different task.

Question 2

Which optimizer is commonly described as using adaptive learning rates per parameter and momentum?

Accepted Answer

Adam

Answer

Adam (Adaptive Moment Estimation) combines momentum and adaptive learning rates, making it a popular default optimizer.

Question 3

What is a residual connection (skip connection) in a neural network?

Accepted Answer

A shortcut that adds the input of a layer block directly to its output

Answer

Residual connections add the block's input directly to its output, enabling training of very deep networks by easing gradient flow.

Question 4

What does the loss function measure in neural network training?

Accepted Answer

The discrepancy between the model's predictions and the true labels

Answer

The loss function quantifies how wrong the model's predictions are, providing the signal that drives parameter updates.

Question 5

Which technique generates new training images by flipping, rotating, or cropping existing ones?

Accepted Answer

Data augmentation

Answer

Data augmentation artificially expands the training set by applying transformations, improving generalization.

Artificial Intelligence Practice Test