CNN Transfer Learning with CNNs 2

Question 1

Which layer type is typically removed when adapting a pretrained CNN for a new classification task?

Accepted Answer

Fully connected output layer

Answer

The final fully connected (classification) layer is replaced because it was trained for the original task's class labels.

Question 2

When fine-tuning only the last few layers of a pretrained CNN, what is the main concern if you use a very high learning rate?

Accepted Answer

Catastrophic forgetting of pretrained features

Answer

A high learning rate can overwrite the valuable pretrained weights through catastrophic forgetting, destroying learned representations.

Question 3

What does 'feature extraction' mode mean in transfer learning?

Accepted Answer

Freezing pretrained layers and only training the new classifier head

Answer

Feature extraction freezes the convolutional base and trains only the added classification head on the new dataset.

Question 4

A pretrained CNN trained on ImageNet is being adapted for satellite imagery classification. Which transfer learning strategy is MOST appropriate?

Accepted Answer

Fine-tune all layers with a small learning rate

Answer

Satellite imagery differs significantly from ImageNet photos, so fine-tuning all layers (with a small LR) allows the model to adapt its features.

Question 5

Which metric best indicates whether a source domain is suitable for transfer to a target domain?

Accepted Answer

Domain similarity or distribution overlap

Answer

Domain similarity determines how transferable the learned features are; high similarity means better transfer.

CNN - Convolutional Neural Networks Practice Test