Research map / Optimization / Optimizers

Meta-Learning Priors Using Unrolled Proximal Networks

ICLR 2024 · Optimization › Optimizers · approach: Gradient-based meta-learning theory

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes meta-learning priors built from unrolled proximal networks, aiming to combine interpretability of explicit priors with the performance of learned models for few-shot learning.

Problem

Meta-learning from a limited number of labeled examples relies on prior knowledge from related tasks; existing prior-based or recurrent approaches have strong empirical performance but limited interpretability.

Approach

Model the meta-learned prior via unrolled proximal (optimization) networks so that the prior has an explicit, interpretable structure while its parameters are learned from labeled data across observed tasks.

Experimental setup

Builds on: model-agnostic meta-learning

Closest papers by meaning

• Meta-Learning and Universality: Deep Representations and Gradient Descent can Approximate any Learning Algorithm (ICLR 2018 · similarity 0.88)
• MARS: Meta-learning as Score Matching in the Function Space (ICLR 2023 · similarity 0.88)
• Meta-Learning Universal Priors Using Non-Injective Change of Variables (NeurIPS 2024 · similarity 0.88)
• Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.86)
• First-order ANIL provably learns representations despite overparametrisation (ICLR 2024 · similarity 0.86)
• Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.86)