Specialized Applications of Data Science
Free CompTIA DataX practice — 6 questions on Specialized Applications of Data Science, with explanations. No sign-up. Full 12-question mixed test →
A manufacturing team needs to optimize a production schedule with 200 discrete variables. The objective function is non-differentiable and non-convex, with many local optima caused by machine changeover constraints. Which optimization approach BEST meets these requirements?
Genetic algorithms are derivative-free metaheuristics that handle discrete, non-differentiable, non-convex search spaces with multiple local optima by maintaining a population and using selection, crossover, and mutation to escape local optima.
A team is building a text classifier for clinical notes that must handle rare medical terms, drug names, and misspellings not seen during training, while keeping vocabulary size manageable. Which tokenization approach BEST meets these requirements?
BPE subword tokenization decomposes unseen or rare words into known subword units, drastically reducing out-of-vocabulary issues while keeping the vocabulary compact—ideal for domain-specific rare terms like medical vocabulary.
During non-maximum suppression (NMS) in an object detection pipeline, two bounding boxes predicted for the same object have an IoU of 0.75, and the NMS IoU threshold is set to 0.5. What happens to these two boxes?
NMS compares IoU against a threshold; when overlap exceeds that threshold (0.75 > 0.5), the box with the lower confidence score is discarded as a duplicate detection, keeping only the highest-confidence box for that object.
A streaming platform launches a new service with zero historical interaction data for brand-new users, but it has rich content metadata including genre, cast, and director for every title. Which recommendation approach BEST meets these requirements for serving new users?
Content-based filtering builds recommendations from item attributes (genre, cast, director) and can generate relevant suggestions for new users immediately, without requiring any prior interaction history—directly solving the user cold-start problem.
Which statement correctly differentiates Isolation Forest from Local Outlier Factor (LOF) for anomaly detection?
Isolation Forest isolates points via random recursive feature splits, using average path length to a leaf as the anomaly score, with no distance computation. LOF instead compares each point's local density to that of its k-nearest neighbors, flagging points in sparser regions as anomalies.
A team needs to tune 8 continuous hyperparameters for a deep learning model where each training run takes 6 hours to complete. Which optimization strategy BEST meets these requirements to minimize total tuning time while finding near-optimal settings?
Bayesian optimization builds a probabilistic surrogate model of the objective and uses an acquisition function to intelligently select the next most promising configuration to evaluate, making it highly sample-efficient—critical when each evaluation costs 6 hours of compute.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →