Matches in SemOpenAlex for { <https://semopenalex.org/work/W3042103612> ?p ?o ?g. }
- W3042103612 abstract "As a popular meta-learning approach, the model-agnostic meta-learning (MAML) algorithm has been widely used due to its simplicity and effectiveness. However, the convergence of the general multi-step MAML still remains unexplored. In this paper, we develop a new theoretical framework to provide such convergence guarantee for two types of objective functions that are of interest in practice: (a) resampling case (e.g., reinforcement learning), where loss functions take the form in expectation and new data are sampled as the algorithm runs; and (b) finite-sum case (e.g., supervised learning), where loss functions take the finite-sum form with given samples. For both cases, we characterize the convergence rate and the computational complexity to attain an $epsilon$-accurate solution for multi-step MAML in the general nonconvex setting. In particular, our results suggest that an inner-stage stepsize needs to be chosen inversely proportional to the number $N$ of inner-stage steps in order for $N$-step MAML to have guaranteed convergence. From the technical perspective, we develop novel techniques to deal with the nested structure of the meta gradient for multi-step MAML, which can be of independent interest." @default.
- W3042103612 created "2020-07-16" @default.
- W3042103612 creator A5042048777 @default.
- W3042103612 creator A5071973105 @default.
- W3042103612 creator A5089183792 @default.
- W3042103612 date "2020-02-18" @default.
- W3042103612 modified "2023-10-18" @default.
- W3042103612 title "Theoretical Convergence of Multi-Step Model-Agnostic Meta-Learning" @default.
- W3042103612 cites W1522301498 @default.
- W3042103612 cites W1771410628 @default.
- W3042103612 cites W2119717200 @default.
- W3042103612 cites W2123399796 @default.
- W3042103612 cites W2138623838 @default.
- W3042103612 cites W2140804329 @default.
- W3042103612 cites W2472819217 @default.
- W3042103612 cites W2601450892 @default.
- W3042103612 cites W2604763608 @default.
- W3042103612 cites W2742093937 @default.
- W3042103612 cites W2753160622 @default.
- W3042103612 cites W2788629937 @default.
- W3042103612 cites W2790068046 @default.
- W3042103612 cites W2794363191 @default.
- W3042103612 cites W2795900505 @default.
- W3042103612 cites W2890870050 @default.
- W3042103612 cites W2895794056 @default.
- W3042103612 cites W2903703072 @default.
- W3042103612 cites W2923966100 @default.
- W3042103612 cites W2925446857 @default.
- W3042103612 cites W2948974578 @default.
- W3042103612 cites W2952003143 @default.
- W3042103612 cites W2952951948 @default.
- W3042103612 cites W2962732055 @default.
- W3042103612 cites W2963303956 @default.
- W3042103612 cites W2963341924 @default.
- W3042103612 cites W2963371846 @default.
- W3042103612 cites W2963547174 @default.
- W3042103612 cites W2963581679 @default.
- W3042103612 cites W2963649256 @default.
- W3042103612 cites W2963887494 @default.
- W3042103612 cites W2964078140 @default.
- W3042103612 cites W2964588180 @default.
- W3042103612 cites W2970389519 @default.
- W3042103612 cites W2994684563 @default.
- W3042103612 cites W2995049146 @default.
- W3042103612 cites W2995627493 @default.
- W3042103612 cites W3005787005 @default.
- W3042103612 cites W3006977545 @default.
- W3042103612 cites W3007684729 @default.
- W3042103612 cites W3033474219 @default.
- W3042103612 cites W3033828374 @default.
- W3042103612 cites W3034334460 @default.
- W3042103612 cites W3035180992 @default.
- W3042103612 cites W3035840208 @default.
- W3042103612 cites W3037059960 @default.
- W3042103612 cites W3037724881 @default.
- W3042103612 cites W3091905774 @default.
- W3042103612 cites W3092849863 @default.
- W3042103612 cites W3101367184 @default.
- W3042103612 cites W3103182070 @default.
- W3042103612 cites W99485931 @default.
- W3042103612 doi "https://doi.org/10.48550/arxiv.2002.07836" @default.
- W3042103612 hasPublicationYear "2020" @default.
- W3042103612 type Work @default.
- W3042103612 sameAs 3042103612 @default.
- W3042103612 citedByCount "13" @default.
- W3042103612 countsByYear W30421036122020 @default.
- W3042103612 countsByYear W30421036122021 @default.
- W3042103612 crossrefType "posted-content" @default.
- W3042103612 hasAuthorship W3042103612A5042048777 @default.
- W3042103612 hasAuthorship W3042103612A5071973105 @default.
- W3042103612 hasAuthorship W3042103612A5089183792 @default.
- W3042103612 hasBestOaLocation W30421036121 @default.
- W3042103612 hasConcept C111472728 @default.
- W3042103612 hasConcept C11413529 @default.
- W3042103612 hasConcept C119857082 @default.
- W3042103612 hasConcept C126255220 @default.
- W3042103612 hasConcept C12713177 @default.
- W3042103612 hasConcept C138885662 @default.
- W3042103612 hasConcept C150921843 @default.
- W3042103612 hasConcept C154945302 @default.
- W3042103612 hasConcept C162324750 @default.
- W3042103612 hasConcept C187736073 @default.
- W3042103612 hasConcept C26517878 @default.
- W3042103612 hasConcept C2776372474 @default.
- W3042103612 hasConcept C2777303404 @default.
- W3042103612 hasConcept C2780451532 @default.
- W3042103612 hasConcept C2781002164 @default.
- W3042103612 hasConcept C33923547 @default.
- W3042103612 hasConcept C38652104 @default.
- W3042103612 hasConcept C41008148 @default.
- W3042103612 hasConcept C50522688 @default.
- W3042103612 hasConcept C57869625 @default.
- W3042103612 hasConcept C97541855 @default.
- W3042103612 hasConceptScore W3042103612C111472728 @default.
- W3042103612 hasConceptScore W3042103612C11413529 @default.
- W3042103612 hasConceptScore W3042103612C119857082 @default.
- W3042103612 hasConceptScore W3042103612C126255220 @default.
- W3042103612 hasConceptScore W3042103612C12713177 @default.
- W3042103612 hasConceptScore W3042103612C138885662 @default.
- W3042103612 hasConceptScore W3042103612C150921843 @default.