Deep Learning Advanced Quiz 2

1. Which technique helps train very deep networks without severe vanishing gradients?

2. What is the “self-attention” mechanism used for?

3. Which method is often used for hyperparameter optimization in deep learning?

4. What is the main function of the encoder in an encoder–decoder architecture?

5. Which activation function can help reduce the dying ReLU problem?

6. What is the main purpose of a good weight initialization strategy?

7. Which architecture is designed for sequential data with long-term dependencies?

8. What is the main advantage of depthwise separable convolutions?

9. Which technique is commonly used for unsupervised representation learning?

10. What is transfer learning?