Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Convergence of Meta-Learning with Task-Specific Adaptation over Partial Parameters

Authors: Kaiyi Ji, Jason D. Lee, Yingbin Liang, H. Vincent Poor

NeurIPS 2020 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experiments on standard few-shot meta-learning benchmarks validate our theoretical findings.
Researcher Affiliation Academia Kaiyi Ji Department of ECE The Ohio State University EMAIL Jason D. Lee Department of EE Princeton University EMAIL Yingbin Liang Department of ECE The Ohio State University EMAIL H. Vincent Poor Department of EE Princeton University EMAIL
Pseudocode Yes Algorithm 1 ANIL Algorithm
Open Source Code No The paper states: 'The experimental implementation and the model architectures are adapted from the existing repository [1] for ANIL.' Reference [1] is 'learn2learn, 2019. https://github.com/learnables/learn2learn.' This indicates use of an existing library, not the release of the authors' own specific implementation code for the methodology described in the paper.
Open Datasets Yes In this section, we validate our theory on the ANIL algorithm over two benchmarks for few-shot multiclass classification, i.e., FC100 [23] and mini Image Net [30].
Dataset Splits No The paper states: 'We consider a 5-way 5-shot task on both the FC100 and mini Image Net datasets.' and 'We relegate the introduction of datasets, model architectures and hyper-parameter settings to Appendix A due to the space limitations.' However, the main text provided does not specify the train/validation/test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not specify any hardware details (e.g., GPU model, CPU type, memory) used for running the experiments.
Software Dependencies No The paper states: 'The experimental implementation and the model architectures are adapted from the existing repository [1] for ANIL.' While reference [1] points to 'learn2learn' on GitHub, no specific version numbers for 'learn2learn' or any other software dependencies (e.g., Python, PyTorch) are provided.
Experiment Setup Yes We consider a 5-way 5-shot task on both the FC100 and mini Image Net datasets. We relegate the introduction of datasets, model architectures and hyper-parameter settings to Appendix A due to the space limitations.