Machine Learning and Neural Networks AIE Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A data scientist is working with a dataset of images. Each image is 256x256 pixels with 3 color channels, resulting in 196,608 features per image. They want to reduce the dimensionality while preserving the most important visual information for a downstream classification task. Which approach is most appropriate?

   • AUse a supervised classifier to select the most important pixels
   • BUse a convolutional autoencoder to learn a compressed representation of the images
   • CApply PCA directly to the raw pixel values
   • DUse t-SNE to reduce the images to 2 dimensions for visualization
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   Which machine learning paradigm uses a reward signal to guide an agent's decisions over time?

   • AReinforcement learning
   • BUnsupervised learning
   • CSemi-supervised learning
   • DSupervised learning
   Reveal Solution Discussion Select an answer first
4. 3expert · hard

   A data science team has a dataset with 500 features and 10,000 unlabeled samples. They need to reduce the feature space to 50 dimensions for a downstream clustering task, but they also want to retain as much variance as possible. They are considering PCA and an autoencoder. Which approach is more appropriate given the constraints?

   • AUse t-SNE, because it preserves local structure and is ideal for reducing to 50 dimensions
   • BUse PCA, because it is a linear method that maximizes variance and is computationally efficient for this dataset size
   • CUse a supervised classifier to select the 50 most important features
   • DUse an autoencoder, because it can capture non-linear relationships and is always better than PCA
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A retail company wants to predict whether a customer will churn within the next 30 days. They have historical data with customer demographics, purchase history, and a binary 'churned' label for past customers. The data scientist needs to build a model that learns the pattern from these labeled examples to classify new customers. Which machine learning approach should they use?

   • ASupervised regression to predict a continuous churn probability score
   • BReinforcement learning where the agent learns a churn-prediction policy from customer interactions
   • CSupervised classification using a neural network with a sigmoid output layer
   • DUnsupervised clustering to group similar customers and infer churn risk from cluster membership
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which component of reinforcement learning is the external system that the agent interacts with and receives feedback from?

   • AAgent
   • BEnvironment
   • CReward function
   • DPolicy
   Reveal Solution Discussion Select an answer first