no code implementations • EMNLP 2020 • Runzhi Tian, Yongyi Mao, Richong Zhang
The introduction of VAE provides an efficient framework for the learning of generative models, including generative topic models.
no code implementations • Findings (ACL) 2022 • Kai Sun, Richong Zhang, Samuel Mensah, Yongyi Mao, Xudong Liu
As like previous work, we rely on negative entities to encourage our model to discriminate the golden entities during training.
no code implementations • 2 Feb 2024 • Ziqiao Wang, Yongyi Mao
Unsupervised domain adaptation (UDA) plays a crucial role in addressing distribution shifts in machine learning.
no code implementations • 28 Nov 2023 • Runzhi Tian, Yongyi Mao
Adversarial training may be regarded as standard training with a modified loss function.
1 code implementation • 28 Oct 2023 • Chonggang Lu, Richong Zhang, Kai Sun, Jaein Kim, Cunwang Zhang, Yongyi Mao
Existing methods focus on building a heterogeneous document graph to model the internal structure of an entity and the external interaction between entities.
1 code implementation • 12 Sep 2023 • Mingxin Li, Richong Zhang, Zhijie Nie, Yongyi Mao
An intriguing phenomenon in CSE is the significant performance gap between supervised and unsupervised methods, with their only difference lying in the training data.
no code implementations • 23 May 2023 • Zhiyi Dong, Yongyi Mao
Building upon Randomized Discretization, we develop two novel adversarial defenses against white-box PGD attacks, utilizing vector quantization in higher dimensional spaces.
1 code implementation • 16 May 2023 • Junfan Chen, Richong Zhang, Yongyi Mao, Jie Xu
Few-shot text classification has recently been promoted by the meta-learning paradigm which aims to identify target classes with knowledge transferred from source classes with sets of small tasks named episodes.
1 code implementation • 16 May 2023 • Junfan Chen, Richong Zhang, Zheyan Luo, Chunming Hu, Yongyi Mao
Data augmentation is widely used in text classification, especially in the low-resource regime where a few examples for each class are available during training.
no code implementations • 2 Mar 2023 • Zixuan Liu, Ziqiao Wang, Hongyu Guo, Yongyi Mao
Mixup, which creates synthetic training instances by linearly interpolating random sample pairs, is a simple and yet effective regularization technique to boost the performance of deep models trained with SGD.
no code implementations • 28 Feb 2023 • Kai Sun, Richong Zhang, Samuel Mensah, Nikolaos Aletras, Yongyi Mao, Xudong Liu
Inspired by the theoretical foundations in domain adaptation [2], we propose a new SSL approach that opts for selecting target samples whose model output from a domain-specific teacher and student network disagree on the unlabelled target data, in an effort to boost the target domain performance.
1 code implementation • 5 Feb 2023 • Ziqiao Wang, Yongyi Mao
In this work, we present a variety of novel information-theoretic generalization bounds for learning algorithms, from the supersample setting of Steinke & Zakynthinou (2020)-the setting of the "conditional mutual information" framework.
no code implementations • 19 Nov 2022 • Ziqiao Wang, Yongyi Mao
Using this estimate, we apply the PAC-Bayes-like information-theoretic bounds developed in both Xu & Raginsky (2017) and Negrea et al. (2019) to obtain generalization upper bounds in terms of the KL divergence between the steady-state weight distribution of SGD with respect to a prior distribution.
no code implementations • 3 Oct 2022 • Ziqiao Wang, Yongyi Mao
This paper uses information-theoretic tools to analyze the generalization error in unsupervised domain adaptation (UDA).
no code implementations • 11 Feb 2022 • Jirui Qi, Richong Zhang, Chune Li, Yongyi Mao
Few-shot relation classification (RC) is one of the critical problems in machine learning.
cross-domain few-shot learning Few-Shot Relation Classification
2 code implementations • 21 Jan 2022 • Qianben Chen, Richong Zhang, Yaowei Zheng, Yongyi Mao
Contrastive learning has achieved remarkable success in representation learning via self-supervision in unsupervised settings.
Ranked #1 on Subjectivity Analysis on SUBJ
no code implementations • 18 Oct 2021 • Hongyu Guo, Yongyi Mao
We present a simple and yet effective interpolation-based regularization technique, aiming to improve the generalization of Graph Neural Networks (GNNs) on supervised graph classification.
no code implementations • ICLR 2022 • Ziqiao Wang, Yongyi Mao
This paper follows up on a recent work of Neu et al. (2021) and presents some new information-theoretic upper bounds for the generalization error of machine learning models, such as neural networks, trained with SGD.
no code implementations • 29 Sep 2021 • Hongyu Guo, Yongyi Mao
We present a simple and yet effective interpolation-based regularization technique to improve the generalization of Graph Neural Networks (GNNs).
no code implementations • 28 May 2021 • Xiaohui Guo, Richong Zhang, Yaowei Zheng, Yongyi Mao
The ALPS regularization objective is formulated as a min-max problem, in which the outer problem is minimizing an upper-bound of the VRM loss, and the inner problem is L$_1$-ball constrained adversarial labelling on perturbed sample.
no code implementations • NeurIPS 2021 • Haoye Lu, Yongyi Mao, Maia Fraser
In particular, we describe a convex optimization problem that arises in a family of estimation tasks commonly appearing in the design of deep learning models.
no code implementations • NeurIPS 2021 • Ziqiao Wang, Yongyi Mao
Understanding the generalization behaviour of deep neural networks is an important theme of modern research in machine learning.
1 code implementation • ICLR 2021 • Haoye Lu, Yongyi Mao, Amiya Nayak
The attention mechanism has been widely used in deep neural networks as a model component.
1 code implementation • CVPR 2021 • Yaowei Zheng, Richong Zhang, Yongyi Mao
This work proposes a new regularization scheme, based on the understanding that the flat local minima of the empirical risk cause the model to generalize better.
Ranked #8 on Image Classification on SVHN
1 code implementation • EMNLP 2020 • Junfan Chen, Richong Zhang, Yongyi Mao, Jie Xu
In this study, we argue that the incorporation of these dependencies is crucial for the design of MDST and propose Parallel Interactive Networks (PIN) to model these dependencies.
Dialogue State Tracking Multi-domain Dialogue State Tracking
1 code implementation • Findings of the Association for Computational Linguistics 2020 • Junfan Chen, Richong Zhang, Yongyi Mao, Jie Xu
Existing DST models either ignore temporal feature dependencies across dialogue turns or fail to explicitly model temporal state dependencies in a dialogue.
no code implementations • 2 Sep 2020 • Ziqiao Wang, Yongyi Mao, Hongyu Guo, Richong Zhang
SkipGram word embedding models with negative sampling, or SGN in short, is an elegant family of word embedding models.
no code implementations • EMNLP 2020 • Kai Sun, Richong Zhang, Samuel Mensah, Yongyi Mao, Xudong Liu
The idea of using multi-task learning approaches to address the joint extraction of entity and relation is motivated by the relatedness between the entity recognition task and the relation classification task.
Ranked #8 on Relation Extraction on WebNLG
1 code implementation • 12 Jan 2020 • Masoumeh Soflaei, Hongyu Guo, Ali Al-Bashabsheh, Yongyi Mao, Richong Zhang
We show that IB learning is, in fact, equivalent to a special class of the quantization problem.
no code implementations • IJCNLP 2019 • Kai Sun, Richong Zhang, Samuel Mensah, Yongyi Mao, Xudong Liu
We propose a method based on neural networks to identify the sentiment polarity of opinion words expressed on a specific aspect of a sentence.
1 code implementation • IJCNLP 2019 • Junfan Chen, Richong Zhang, Yongyi Mao, Hongyu Guo, Jie Xu
Distant supervision for relation extraction enables one to effectively acquire structured relations out of very large text corpora with less human efforts.
no code implementations • ICLR 2020 • Guillaume P. Archambault, Yongyi Mao, Hongyu Guo, Richong Zhang
We prove that the family of Untied MixUp schemes is equivalent to the entire class of DAT schemes.
2 code implementations • 22 May 2019 • Hongyu Guo, Yongyi Mao, Richong Zhang
Mixup, a recent proposed data augmentation method through linearly interpolating inputs and modeling targets of random samples, has demonstrated its capability of significantly improving the predictive accuracy of the state-of-the-art networks for image classification.
no code implementations • 3 Mar 2019 • Bokang Zhu, Richong Zhang, Dingkun Long, Yongyi Mao
Gated models resolve this conflict by adaptively adjusting their state-update equations, whereas Vanilla RNN resolves this conflict by assigning different dimensions different tasks.
no code implementations • EMNLP 2018 • Richong Zhang, Zhiyuan Hu, Hongyu Guo, Yongyi Mao
We propose a novel strategy to encode the syntax parse tree of sentence into a learnable distributed representation.
2 code implementations • 7 Sep 2018 • Hongyu Guo, Yongyi Mao, Richong Zhang
To address this issue, we propose a novel adaptive version of MixUp, where the mixing policies are automatically learned from the data using an additional network and objective function designed to avoid manifold intrusion.
no code implementations • COLING 2018 • Yue Wang, Richong Zhang, Cheng Xu, Yongyi Mao
In this paper, we study the problem of question answering over knowledge base.
no code implementations • 26 Jul 2018 • Hongyu Guo, Yongyi Mao, Ali Al-Bashabsheh, Richong Zhang
Based on the notion of information bottleneck (IB), we formulate a quantization problem called "IB quantization".
no code implementations • 20 Nov 2016 • Dingkun Long, Richong Zhang, Yongyi Mao
For this purpose, we design a simple and controllable task, called ``memorization problem'', where the networks are trained to memorize certain targeted information.
no code implementations • 28 Apr 2016 • Jianfeng Wen, Jian-Xin Li, Yongyi Mao, Shini Chen, Richong Zhang
The models developed to date for knowledge base embedding are all based on the assumption that the relations contained in knowledge bases are binary.