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

First convolutional layer

Answer

Batch normalization layer

Answer

Max pooling layer

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

Underfitting the new dataset

Answer

Overfitting the training data

Answer

Increased inference time

Question 3

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

Accepted Answer

Freezing pretrained layers and only training the new classifier head

Answer

Training all layers from scratch on new data

Answer

Extracting raw pixel features before convolution

Answer

Using PCA to reduce feature dimensions

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

Use the model as-is with no retraining

Answer

Freeze all layers and only add a new softmax output

Answer

Train from scratch ignoring ImageNet weights

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

Number of parameters in the source model

Answer

Depth of the source CNN

Answer

Batch size used during source training

Question 6

In progressive fine-tuning, layers are unfrozen from which direction?

Accepted Answer

From the last (output) layer backward

Answer

From the first (input) layer forward

Answer

All layers simultaneously

Answer

Only middle layers first

Question 7

What is 'negative transfer' in the context of transfer learning with CNNs?

Accepted Answer

When pretraining hurts performance on the target task

Answer

Transfer learning that reduces model size

Answer

Transferring weights with negative values

Answer

Training on negative examples only