AI AI Engineer: Neural Networks and Deep Learning 2

Question 1

What is a recurrent neural network (RNN) primarily designed to handle?

Accepted Answer

Sequential and time-series data with temporal dependencies

Answer

RNNs maintain a hidden state across time steps, making them suited for sequential data like text, audio, and time series.

Question 2

What problem do LSTM (Long Short-Term Memory) networks solve compared to vanilla RNNs?

Accepted Answer

Long-range dependency learning by mitigating vanishing gradients over long sequences

Answer

LSTMs use gating mechanisms (input, forget, output gates) to selectively retain or discard information, enabling learning of long-range dependencies.

Question 3

In deep learning, what is transfer learning?

Accepted Answer

Reusing a model pretrained on a large dataset as a starting point for a new task

Answer

Transfer learning leverages knowledge from a pretrained model (e.g., ImageNet-trained ResNet) by fine-tuning it on a smaller task-specific dataset.

Question 4

What is the role of the 'softmax' activation function in a multi-class classification output layer?

Accepted Answer

Converts raw logits into a probability distribution summing to 1

Answer

Softmax exponentiates each logit and divides by the sum of all exponentiated logits, producing class probabilities that sum to 1.

Question 5

Which deep learning architecture introduced skip (residual) connections to enable training of very deep networks?

Accepted Answer

ResNet

Answer

ResNet introduced residual connections that add layer inputs directly to outputs, allowing gradients to flow more easily and enabling hundreds of layers.

AI - Engineer Practice Test

AI - Engineer Practice Test

AI AI Engineer: Neural Networks and Deep Learning 2