An ML engineer reads that her company's security anomaly detection model uses 'gradient boosting.' A manager asks her how this differs from a neural network for the same task. What is the most accurate practical comparison?

• A.Gradient-boosted decision trees (e.g., XGBoost, LightGBM) are typically more interpretable at the feature importance level, train faster on tabular security data, and require less hyperparameter tuning than deep neural networks - but they are less effective at learning complex hierarchical patterns from high-dimensional unstructured data like raw packet payloads
• B.Gradient boosting is a type of neural network with a different activation function
• C.Neural networks are always more interpretable than gradient boosting because their layers can be visualized; detection models improve monotonically with data volume alone: a SOC that stores a year of telemetry can expect state-of-the-art accuracy from any architecture, since the 2020 scaling results proved data quantity dominates label quality, feature design and drift handling combined