Matches in SemOpenAlex for { <https://semopenalex.org/work/W2997952181> ?p ?o ?g. }
- W2997952181 endingPage "7520" @default.
- W2997952181 startingPage "7512" @default.
- W2997952181 abstract "Reinforcement learning (RL) has been widely used to aid training in language generation. This is achieved by enhancing standard maximum likelihood objectives with user-specified reward functions that encourage global semantic consistency. We propose a principled approach to address the difficulties associated with RL-based solutions, namely, high-variance gradients, uninformative rewards and brittle training. By leveraging the optimal transport distance, we introduce a regularizer that significantly alleviates the above issues. Our formulation emphasizes the preservation of semantic features, enabling end-to-end training instead of ad-hoc fine-tuning, and when combined with RL, it controls the exploration space for more efficient model updates. To validate the effectiveness of the proposed solution, we perform a comprehensive evaluation covering a wide variety of NLP tasks: machine translation, abstractive text summarization and image caption, with consistent improvements over competing solutions." @default.
- W2997952181 created "2020-01-10" @default.
- W2997952181 creator A5016448581 @default.
- W2997952181 creator A5023554048 @default.
- W2997952181 creator A5029042665 @default.
- W2997952181 creator A5031220156 @default.
- W2997952181 creator A5056639842 @default.
- W2997952181 creator A5058762520 @default.
- W2997952181 creator A5071582087 @default.
- W2997952181 creator A5081872106 @default.
- W2997952181 date "2020-04-03" @default.
- W2997952181 modified "2023-09-26" @default.
- W2997952181 title "Sequence Generation with Optimal-Transport-Enhanced Reinforcement Learning" @default.
- W2997952181 cites W1514535095 @default.
- W2997952181 cites W1544827683 @default.
- W2997952181 cites W1771410628 @default.
- W2997952181 cites W1861492603 @default.
- W2997952181 cites W1895577753 @default.
- W2997952181 cites W1905882502 @default.
- W2997952181 cites W1956340063 @default.
- W2997952181 cites W2099471712 @default.
- W2997952181 cites W2101105183 @default.
- W2997952181 cites W2119717200 @default.
- W2997952181 cites W2123301721 @default.
- W2997952181 cites W2130942839 @default.
- W2997952181 cites W2143668817 @default.
- W2997952181 cites W2157331557 @default.
- W2997952181 cites W2158131535 @default.
- W2997952181 cites W2302086703 @default.
- W2997952181 cites W2467173223 @default.
- W2997952181 cites W2508728158 @default.
- W2997952181 cites W2525778437 @default.
- W2997952181 cites W2547875792 @default.
- W2997952181 cites W2558834163 @default.
- W2997952181 cites W2574535369 @default.
- W2997952181 cites W2575842049 @default.
- W2997952181 cites W2606974598 @default.
- W2997952181 cites W2612675303 @default.
- W2997952181 cites W2739748921 @default.
- W2997952181 cites W2745461083 @default.
- W2997952181 cites W2786002645 @default.
- W2997952181 cites W2797269287 @default.
- W2997952181 cites W2803930360 @default.
- W2997952181 cites W2889518897 @default.
- W2997952181 cites W2891641674 @default.
- W2997952181 cites W2899771611 @default.
- W2997952181 cites W2910092637 @default.
- W2997952181 cites W2951493172 @default.
- W2997952181 cites W2952165242 @default.
- W2997952181 cites W2962965405 @default.
- W2997952181 cites W2962985882 @default.
- W2997952181 cites W2963033554 @default.
- W2997952181 cites W2963084599 @default.
- W2997952181 cites W2963248296 @default.
- W2997952181 cites W2963366196 @default.
- W2997952181 cites W2963403868 @default.
- W2997952181 cites W2963545005 @default.
- W2997952181 cites W2963631950 @default.
- W2997952181 cites W2963673305 @default.
- W2997952181 cites W2963929190 @default.
- W2997952181 cites W2963989027 @default.
- W2997952181 cites W2964017345 @default.
- W2997952181 cites W2964201867 @default.
- W2997952181 cites W2964268978 @default.
- W2997952181 cites W2964308564 @default.
- W2997952181 cites W3041866211 @default.
- W2997952181 cites W3145992521 @default.
- W2997952181 cites W639708223 @default.
- W2997952181 cites W648786980 @default.
- W2997952181 cites W658020064 @default.
- W2997952181 doi "https://doi.org/10.1609/aaai.v34i05.6249" @default.
- W2997952181 hasPublicationYear "2020" @default.
- W2997952181 type Work @default.
- W2997952181 sameAs 2997952181 @default.
- W2997952181 citedByCount "9" @default.
- W2997952181 countsByYear W29979521812020 @default.
- W2997952181 countsByYear W29979521812021 @default.
- W2997952181 countsByYear W29979521812023 @default.
- W2997952181 crossrefType "journal-article" @default.
- W2997952181 hasAuthorship W2997952181A5016448581 @default.
- W2997952181 hasAuthorship W2997952181A5023554048 @default.
- W2997952181 hasAuthorship W2997952181A5029042665 @default.
- W2997952181 hasAuthorship W2997952181A5031220156 @default.
- W2997952181 hasAuthorship W2997952181A5056639842 @default.
- W2997952181 hasAuthorship W2997952181A5058762520 @default.
- W2997952181 hasAuthorship W2997952181A5071582087 @default.
- W2997952181 hasAuthorship W2997952181A5081872106 @default.
- W2997952181 hasBestOaLocation W29979521811 @default.
- W2997952181 hasConcept C111919701 @default.
- W2997952181 hasConcept C119857082 @default.
- W2997952181 hasConcept C121955636 @default.
- W2997952181 hasConcept C136197465 @default.
- W2997952181 hasConcept C144133560 @default.
- W2997952181 hasConcept C154945302 @default.
- W2997952181 hasConcept C170858558 @default.
- W2997952181 hasConcept C196083921 @default.
- W2997952181 hasConcept C204321447 @default.
- W2997952181 hasConcept C2776436953 @default.