Deep Learning DY0-001 Practice Questions (Page 5)

Questions 21–24

1. 21application · easy

   A neural network for sentiment analysis is overfitting. The team decides to add dropout to the hidden layers. What is the primary effect of dropout during training?

   • AIt reduces the learning rate during training to improve convergence.
   • BIt randomly deactivates a fraction of neurons in each forward pass.
   • CIt adds a penalty term to the loss function based on the magnitude of weights.
   • DIt normalizes the inputs to each layer to have zero mean and unit variance.
   Reveal Solution Discussion Select an answer first
2. 22application · easy

   A startup is building a recommendation system and needs to prototype a neural network quickly. The team has limited GPU resources and wants to use a high-level API that abstracts away low-level details. Which framework is most appropriate for this scenario?

   • APyTorch with manual gradient computation
   • BPyTorch with custom autograd functions
   • CTensorFlow with the Estimator API
   • DTensorFlow with the Keras API
   Reveal Solution Discussion Select an answer first
4. 23application · easy

   A machine learning engineer is using TensorFlow to train a neural network. They want to compute gradients and update weights automatically. Which TensorFlow feature should they use?

   • Atf.data.Dataset
   • Btf.keras.optimizers.Adam
   • Ctf.GradientTape
   • Dtf.keras.layers.Dense
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   During which phase of neural network training is dropout typically applied?

   • AOnly during the forward pass of training, not during inference
   • BOnly during the backward pass of training, not during the forward pass
   • CDuring both training and inference to ensure consistent behavior
   • DOnly during validation to evaluate the model's robustness
   Reveal Solution Discussion Select an answer first