Agentic AI Reference Architecture

Model Adaptation · Software component

DAgger Trainer

Software componentModel AdaptationModelsarc:DAggerTrainer

An imitation learner that iteratively executes its current policy, has an expert label the visited states, aggregates these labels into the dataset, and retrains.

Responsibility. Aggregates expert labels for policy-visited states and retrains iteratively.

Also known as: Dataset Aggregation, DAgger

Variant of Imitation Learner abstract

When to choose. Choose when experts can provide repeated labels and the learned policy can be safely executed during training to expose its failure modes.

reads; writessends data toinvokesalternative tospecializesis target of alternativeToAgent Trajectory Dataset: reads; writesAgent Trajectory DatasetDomain Expert Annotator: sends data toDomain Expert AnnotatorEnvironment Simulator: invokesEnvironment SimulatorInverse RL Reward Learner: alternative toInverse RL Reward LearnerImitation Learner: specializesImitation LearnerBehavior Cloning Trainer: is target of alternativeToBehavior Cloning Trainer
Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

reads dependency

writes dependency

sends data to dynamic

alternative to variability

Design guidance

Quantitative guidance

As stated by the sources; verify before use.

Classification

Patterns
Dataset aggregationBehavior-cloning initialization
Risks mitigated
Distribution shift / compounding errors in behavior cloning

Sources

  1. Ch5.12: T. Nguyen, "Learning-Based Decision Making Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.12. ISBN: 9798244538229.