Filters








1,340 Hits in 10.6 sec

MetaMT, a Meta Learning Method Leveraging Multiple Domain Data for Low Resource Machine Translation

Rumeng Li, Xun Wang, Hong Yu
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
In this paper, we present a novel NMT model with a new word embedding transition technique for fast domain adaption.  ...  We mimic the domain adaptation of the machine translation model to low-resource domains using multiple translation tasks on different domains.  ...  Conclusion We present a meta learning method for neural machine translation with limited resources.  ... 
doi:10.1609/aaai.v34i05.6339 pmid:34094698 pmcid:PMC8174838 fatcat:wtwamh3e6bdv3hmv4b74jp53wa

MetaMT,a MetaLearning Method Leveraging Multiple Domain Data for Low Resource Machine Translation [article]

Rumeng Li, Xun Wang, Hong Yu
2019 arXiv   pre-print
Manipulating training data leads to robust neural models for MT.  ...  Conclusion We present a meta learning method for neural machine translation with limited resources.  ...  Curriculum learning has also been exploited (Zhang et al. 2019 ) for domain adaptation.  ... 
arXiv:1912.05467v1 fatcat:zkyfn2woingl3laltyrdle5efm

Improving both domain robustness and domain adaptability in machine translation [article]

Wen Lai, Jindřich Libovický, Alexander Fraser
2022 arXiv   pre-print
In this paper, we propose a novel approach, RMLNMT (Robust Meta-Learning Framework for Neural Machine Translation Domain Adaptation), which improves the robustness of existing meta-learning models.  ...  We consider two problems of NMT domain adaptation using meta-learning.  ...  We thank Mengjie Zhao for the helpful comments.  ... 
arXiv:2112.08288v2 fatcat:yjcvaqjhd5bqpmof4dooygdmpa

Reinforcement Learning based Curriculum Optimization for Neural Machine Translation [article]

Gaurav Kumar, George Foster, Colin Cherry, Maxim Krikun
2019 arXiv   pre-print
We consider the problem of making efficient use of heterogeneous training data in neural machine translation (NMT).  ...  Rather than relying on prior knowledge to design a curriculum, we use reinforcement learning to learn one automatically, jointly with the NMT system, in the course of a single training run.  ...  Conclusion We have presented a method to learn the curriculum for presenting training samples to an NMT system.  ... 
arXiv:1903.00041v1 fatcat:qczttdjsnjhz7hlirdmfgmt7c4

Curriculum Learning: A Survey [article]

Petru Soviany, Radu Tudor Ionescu, Paolo Rota, Nicu Sebe
2022 arXiv   pre-print
In this survey, we show how these limits have been tackled in the literature, and we present different curriculum learning instantiations for various tasks in machine learning.  ...  Curriculum learning strategies have been successfully employed in all areas of machine learning, in a wide range of tasks.  ...  Acknowledgements The authors would like to thank the reviewers for their useful feedback.  ... 
arXiv:2101.10382v3 fatcat:doognr7ggfaalg7kd2i3n7s3jy

Reinforcement Learning based Curriculum Optimization for Neural Machine Translation

Gaurav Kumar, George Foster, Colin Cherry, Maxim Krikun
2019 Proceedings of the 2019 Conference of the North  
We consider the problem of making efficient use of heterogeneous training data in neural machine translation (NMT).  ...  Rather than relying on prior knowledge to design a curriculum, we use reinforcement learning to learn one automatically, jointly with the NMT system, in the course of a single training run.  ...  Acknowledgements The authors would like to thank Wei Wang for his advice and help in replicating the CDS baselines.  ... 
doi:10.18653/v1/n19-1208 dblp:conf/naacl/KumarFCK19 fatcat:mm6ak5wjjfhplldtndks62ufr4

Unsupervised Curricula for Visual Meta-Reinforcement Learning [article]

Allan Jabri, Kyle Hsu, Ben Eysenbach, Abhishek Gupta, Sergey Levine, Chelsea Finn
2019 arXiv   pre-print
We develop an unsupervised algorithm for inducing an adaptive meta-training task distribution, i.e. an automatic curriculum, by modeling unsupervised interaction in a visual environment.  ...  In experiments on vision-based navigation and manipulation domains, we show that the algorithm allows for unsupervised meta-learning that transfers to downstream tasks specified by hand-crafted reward  ...  Au- tomated curriculum learning for neural networks. In International Conference on Machine Learning (ICML), 2017. [24] Karol Gregor, Danilo Jimenez Rezende, and Daan Wierstra.  ... 
arXiv:1912.04226v1 fatcat:cer4sn5jm5ezrfmcvsvfs3t6ii

Data Selection Curriculum for Neural Machine Translation [article]

Tasnim Mohiuddin, Philipp Koehn, Vishrav Chaudhary, James Cross, Shruti Bhosale, Shafiq Joty
2022 arXiv   pre-print
Neural Machine Translation (NMT) models are typically trained on heterogeneous data that are concatenated and randomly shuffled. However, not all of the training data are equally useful to the model.  ...  In this work, we introduce a two-stage curriculum training framework for NMT where we fine-tune a base NMT model on subsets of data, selected by both deterministic scoring using pre-trained methods and  ...  In recent years, Neural Machine Translation (NMT) has shown impressive performance in highresource settings (Hassan et al., 2018; Popel et al., 2020) .  ... 
arXiv:2203.13867v1 fatcat:qpxundsobzbtvk6fiuwxidivne

Small Sample Learning in Big Data Era [article]

Jun Shu, Zongben Xu, Deyu Meng
2018 arXiv   pre-print
The second category is called "experience learning", which usually co-exists with the large sample learning manner of conventional machine learning.  ...  As a promising area in artificial intelligence, a new learning paradigm, called Small Sample Learning (SSL), has been attracting prominent research attention in the recent years.  ...  Dual learning is firstly propose by (He et al., 2016a) in neural machine translation.  ... 
arXiv:1808.04572v3 fatcat:lqqzzrmgfnfb3izctvdzgopuny

Unsupervised Domain Adaptation for Semantic Image Segmentation: a Comprehensive Survey [article]

Gabriela Csurka, Riccardo Volpi, Boris Chidlovskii
2021 arXiv   pre-print
We present the most important semantic segmentation methods; we provide a comprehensive survey on domain adaptation techniques for semantic segmentation; we unveil newer trends such as multi-domain learning  ...  , domain generalization, test-time adaptation or source-free domain adaptation; we conclude this survey by describing datasets and benchmarks most widely used in semantic segmentation research.  ...  Curriculum Manager for Source Selection in domain adaptation method for semantic segmentation. In Multi-Source Domain Adaptation.  ... 
arXiv:2112.03241v1 fatcat:uzlehddvuvfwzf4dfbjimja45e

Automatically Composing Representation Transformations as a Means for Generalization [article]

Michael B. Chang, Abhishek Gupta, Sergey Levine, Thomas L. Griffiths
2019 arXiv   pre-print
A generally intelligent learner should generalize to more complex tasks than it has previously encountered, but the two common paradigms in machine learning -- either training a separate learner per task  ...  As a first step for tackling compositional generalization, we introduce the compositional recursive learner, a domain-general framework for learning algorithmic procedures for composing representation  ...  Doing more with less: Meta-reasoning and meta-learning in humans and machines. Current Opinion in Behavioral Sciences, 01 2019.Abhishek Gupta, Benjamin Eysenbach, Chelsea Finn, and Sergey Levine.  ... 
arXiv:1807.04640v2 fatcat:rupxorh2lndmpgophzwjzo4a5a

Meta-Learning in Neural Networks: A Survey [article]

Timothy Hospedales, Antreas Antoniou, Paul Micaelli, Amos Storkey
2020 arXiv   pre-print
The field of meta-learning, or learning-to-learn, has seen a dramatic rise in interest in recent years.  ...  Finally, we discuss outstanding challenges and promising areas for future research.  ...  graph extractor [269] , machine translation [270] , and quickly adapting to new personas in dialogue [271] .  ... 
arXiv:2004.05439v2 fatcat:3r23tsxxkfbgzamow5miglkrye

A Survey on Curriculum Learning [article]

Xin Wang, Yudong Chen, Wenwu Zhu
2021 arXiv   pre-print
Curriculum learning (CL) is a training strategy that trains a machine learning model from easier data to harder data, which imitates the meaningful learning order in human curricula.  ...  Finally, we present our insights on the relationships connecting CL and other machine learning concepts including transfer learning, meta-learning, continual learning and active learning, etc., then point  ...  are widely chosen to evaluate CL methods for neural machine translation with BLEU metric [53] , [68] , [86] .  ... 
arXiv:2010.13166v2 fatcat:j5xr3zpgojhannawbdng76ejxy

Meta-Learning in Neural Networks: A Survey

Timothy M Hospedales, Antreas Antoniou, Paul Micaelli, Amos J. Storkey
2021 IEEE Transactions on Pattern Analysis and Machine Intelligence  
The field of meta-learning, or learning-to-learn, has seen a dramatic rise in interest in recent years.  ...  In this survey we describe the contemporary meta-learning landscape.  ...  graph extractor [266] , machine translation [267] , and quickly adapting to new personas in dialogue [268] .  ... 
doi:10.1109/tpami.2021.3079209 pmid:33974543 fatcat:wkzeodki4fbcnjlcczn4mr6kry

A Survey of Domain Adaptation for Machine Translation

Chenhui Chu, Rui Wang
2020 Journal of Information Processing  
Neural machine translation (NMT) is a deep learning based approach for machine translation, which outperforms traditional statistical machine translation (SMT) and yields the state-of-the-art translation  ...  Domain adaptation that leverages both out-of-domain parallel corpora as well as monolingual corpora for in-domain translation, is very important for domain-specific translation.  ...  Acknowledgments This work was supported by Grant-in-Aid for Young Scientists #19K20343, JSPS and Microsoft Research Asia Collaborative Research Grant. We are very indebted to Dr.  ... 
doi:10.2197/ipsjjip.28.413 fatcat:eeboqsm6rfbu7hd6dr4q6u4o3e
« Previous Showing results 1 — 15 out of 1,340 results