Machine Learning and Neural Networks AIE Practice Questions (Page 4)

Questions 16–20

1. 16expert · medium

   A team is building a deep learning model for speech recognition. They have a large dataset of audio recordings and transcripts. They are considering using a deep neural network (DNN) with many hidden layers. What is a key advantage of using a DNN over a shallow network for this task?

   • ADNNs do not require activation functions
   • BDNNs require less training data than shallow networks
   • CDNNs are guaranteed to converge faster than shallow networks
   • DDNNs can learn hierarchical feature representations, from low-level audio features to high-level linguistic patterns
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A company wants to build a system that recognizes objects in real-time video streams. They plan to use a neural network with many hidden layers. What is the primary advantage of using a deep neural network with multiple hidden layers over a shallow network with a single hidden layer?

   • ADeep networks eliminate the need for activation functions
   • BDeep networks can learn hierarchical features, from simple edges to complex object parts, which improves accuracy on complex tasks
   • CDeep networks require less training data than shallow networks
   • DDeep networks are guaranteed to converge faster than shallow networks
   Reveal Solution Discussion Select an answer first
4. 18application · easy

   A bank wants to predict the exact dollar amount of a loan that a customer is likely to default on, based on their income, credit score, and existing debt. The target variable is a continuous monetary value. Which machine learning task is this?

   • ASupervised classification, because the model classifies customers as risky or safe
   • BReinforcement learning, because the model learns from loan repayment rewards
   • CUnsupervised clustering, because the model groups similar customers
   • DSupervised regression, because the target is a continuous value
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   In supervised learning, what is the role of labeled data?

   • AIt provides a set of rules that the model must follow to make predictions.
   • BIt provides the correct output for each input example, which the model uses to learn the mapping from inputs to outputs.
   • CIt provides only input features without any output values, allowing the model to find hidden structures.
   • DIt provides a reward signal that guides the model's actions in an environment.
   Reveal Solution Discussion Select an answer first
6. 20application · easy

   A data scientist is training a deep neural network for image recognition. After each forward pass, the network computes a loss value comparing its predictions to the true labels. The scientist then adjusts the weights to reduce this loss. What is the correct sequence of operations in this training step?

   • AForward propagation to compute predictions, calculate loss, then backpropagation to update weights
   • BForward propagation to update weights, then backpropagation to compute the loss
   • CCalculate loss first, then forward propagation to adjust weights, and finally backpropagation to validate
   • DBackpropagation to compute predictions, calculate loss, then forward propagation to update weights
   Reveal Solution Discussion Select an answer first