Deep Learning

Question 1

What is a CNN?

Answer 1

A deep learning architecture designed for processing data with grid-like topology, e.g., images.

Question 2

What is an RNN?

Answer 2

A neural network designed for sequential data where outputs depend on previous computations.

Question 3

What is a transformer model?

Answer 3

An architecture relying on self-attention mechanisms, excelling in NLP tasks.

Question 4

What is ReLU?

Answer 4

Activation function: f(x) = max(0, x). Helps avoid vanishing gradients.

Question 5

What is sigmoid activation?

Answer 5

Outputs values in (0,1), often used for binary classification.

Question 6

What is softmax?

Answer 6

Converts logits into probability distribution over classes.

Question 7

What is cross-entropy loss?

Answer 7

A loss function for classification comparing predicted probabilities to true labels.

Question 8

What is dropout?

Answer 8

A regularization method that randomly deactivates neurons during training to prevent overfitting.

Question 9

What is batch normalization?

Answer 9

A technique to normalize inputs to each layer, speeding up training.

Question 10

What is Adam optimizer?

Answer 10

Adaptive Moment Estimation optimizer combining RMSProp and momentum.