1. Question 1

If I put a dropout parameter of 0.2, how many nodes will I lose?

• ✅ 20% of them

• ❌ 2% of them

• ❌ 20% of the untrained ones

• ❌ 2% of the untrained ones

Explanation: Dropout(0.2) randomly disables 20% of neurons during each forward pass in training.

---

2. Question 2

How do you change the number of classes when using transfer learning?

• ❌ Ignore all classes above yours

• ❌ Use all classes but set weights to 0

• ✅ When you add your DNN at the bottom, specify your output layer with the number of classes you want

• ❌ Use dropouts to eliminate classes

Explanation: In transfer learning, the base model stays; you replace the final Dense layer with your own output layer (e.g., Dense(2) for 2 classes).

---

3. Question 3

Which is the correct line of code for declaring dropout of 20%?

• ❌ tf.keras.layers.Dropout(20)

• ❌ tf.keras.layers.DropoutNeurons(20)

• ✅ tf.keras.layers.Dropout(0.2)

• ❌ tf.keras.layers.DropoutNeurons(0.2)

Explanation: Dropout takes a fraction (0–1), not a percentage or neuron count.

---

4. Question 4

Why do dropouts help avoid overfitting?

• ✅ Because neighbor neurons can have similar weights, and thus can skew the final training

• ❌ Having fewer neurons speeds up training

Explanation: Dropout breaks co-adaptation between neurons. They can’t rely on each other, so they learn stronger generalizable features.

---

5. Question 5

Why is transfer learning useful?

• ❌ Use all original training data

• ❌ Use all original validation data

• ✅ Use features learned from large datasets you may not have access to

• ❌ Use validation metadata from large datasets

Explanation: Transfer learning imports pre-learned features from huge datasets (like ImageNet) so you benefit without training from scratch.

---

6. Question 6

Can you use image augmentation with transfer learning?

• ❌ No, because features are pre-set

• ✅ Yes, you can augment when training the layers you added

Explanation: Data augmentation improves the new classifier layers you train.

---

7. Question 7

How do you freeze a layer from retraining?

• ❌ tf.freeze(layer)

• ❌ tf.layer.frozen = True

• ❌ tf.layer.locked = True

• ✅ layer.trainable = False

Explanation: Setting trainable=False tells TensorFlow not to update weights for that layer.

---

8. Question 8

What happens if dropout rate is too high?

• ✅ The network becomes ineffective at learning

• ❌ Training time increases due to extra calculations

Explanation: Too much dropout removes too many neurons, making learning weak and accuracy low.

---

🧾 Summary Table

---