Machine Learning and Neural Networks AIE Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A cybersecurity team wants to detect unusual network traffic that may indicate a new type of attack. They have a large dataset of network flows but no labels indicating which flows are attacks. They need to identify anomalies without prior knowledge of attack signatures. Which approach is most appropriate?

   • AReinforcement learning to learn a policy for blocking traffic
   • BSupervised classification using labeled attack and normal traffic
   • CUnsupervised anomaly detection using clustering or autoencoders
   • DDimensionality reduction only, such as PCA, to visualize the data
   Reveal Solution Discussion Select an answer first
2. 22expert · medium

   A company has a dataset of customer transactions. They want to identify unusual transactions that may indicate fraud. They do not have labeled examples of fraud. Which machine learning approach is most suitable for this anomaly detection task?

   • AUnsupervised learning, using clustering or isolation forest to identify outliers
   • BDimensionality reduction, using PCA to reduce features and then manually inspect the data
   • CReinforcement learning, where the model learns to flag transactions based on rewards
   • DSupervised learning, training a classifier on normal vs. fraudulent transactions
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A team currently uses a rule-based system to detect fraudulent transactions by applying fixed thresholds (e.g., flag transactions above $10,000). Fraud patterns change frequently, and the team must manually update the rules. They want a system that learns from historical transaction data and adapts to new fraud patterns automatically. Which approach best addresses this need?

   • AReplace the rule-based system with a machine learning model trained on labeled historical transactions
   • BUse reinforcement learning to learn fraud patterns from a static dataset
   • CKeep the rule-based system but increase the threshold to reduce false positives
   • DUse a clustering algorithm on historical transactions to identify fraud without labels
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   Which technique is an example of dimensionality reduction in unsupervised learning?

   • ALinear Regression
   • BPrincipal Component Analysis (PCA)
   • CDecision Tree Classification
   • DK-Means Clustering
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A robotics company is developing a warehouse robot that must navigate to pick items and avoid obstacles. The robot receives a reward for successfully picking an item and a penalty for colliding. It learns by trial and error, adjusting its actions based on the rewards received. Which machine learning paradigm is being used?

   • AUnsupervised learning, where the robot clusters sensor readings to find safe paths
   • BReinforcement learning, where the robot is the agent, the warehouse is the environment, and the policy maps states to actions
   • CSupervised learning, where the robot is trained on a dataset of optimal paths
   • DDeep learning with backpropagation, where the robot minimizes a loss function over sensor data
   Reveal Solution Discussion Select an answer first