Deep Learning Intermediate Quiz 5

1. What is the main purpose of a pooling layer in a CNN?

2. What is a confusion matrix used for?

3. Which is a benefit of using pre-trained models?

4. How do one-hot encoding and embeddings mainly differ?

5. What is a key property of softmax?

6. In CNNs, what is an activation map?

7. What is the main function of the output layer in a neural network?

8. Why are callbacks useful during training?

9. What is the main benefit of residual (skip) connections in deep networks?

10. In sequence models, what advantage do bidirectional RNNs provide over unidirectional RNNs?