Meta-Learning
Learning to learn — training AI systems that can quickly adapt to new tasks with minimal data by leveraging experience from previous tasks.
Approaches
MAML (Model-Agnostic Meta-Learning): Learn initial parameters that can be quickly fine-tuned. Prototypical Networks: Learn a metric space for few-shot classification. Memory-augmented: Neural networks with external memory for fast adaptation.
Connection to LLMs
In-context learning (providing examples in the prompt) is a form of meta-learning — the model learned during pre-training how to learn from examples at inference time. This was an unexpected emergent capability of large language models.
Applications
Few-shot image classification (recognizing new objects from one photo). Drug discovery (predicting properties of novel molecules). Robotics (quickly adapting to new tasks). Personalization (adapting to individual user preferences with minimal interaction).