What Is Domain Adaptation in Machine Learning?
Domain adaptation adjusts a model trained on one data distribution so it performs better on a related target distribution.
Don't Just Read it – Explore It
Domain adaptation adjusts a model trained on one data distribution so it performs better on a related target distribution.
Out-of-distribution detection tries to identify inputs that differ meaningfully from the data a model was built to handle.
Conformal prediction wraps a model’s output in a calibrated set or interval with a target long-run coverage guarantee under assumptions.
Aleatoric uncertainty comes from variability or noise, while epistemic uncertainty comes from limited knowledge about the model.
Multi-agent reinforcement learning studies how several learning agents cooperate, compete, or do both in a shared environment.
Model uncertainty describes how unsure a system should be about a prediction because of limited knowledge, data, or model assumptions.
Model-based reinforcement learning uses a known or learned environment model to predict outcomes and improve decisions.
Self-play trains an agent by having it compete or interact with copies and past versions of itself.
A world model learns an internal representation of how an environment changes so an agent can predict and plan.
Direct preference optimization trains a model directly from preferred and rejected responses without a separate reward-model training stage.
A reward model learns to score outputs or actions as a proxy for preferences that are difficult to write as a simple rule.
Low-rank adaptation, or LoRA, fine-tunes a model by training small low-rank weight updates while freezing the original weights.