检索结果-内蒙古大学图书馆

Text Adversarial Purification as Defense against Adversarial Attacks

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Linyang Song, Demin Qiu, Xipeng School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

Adversarial purification is a successful defense mechanism against adversarial attacks without requiring knowledge of the form of the incoming attack. Generally, adversarial purification aims to remove the adversarial perturbations therefore can make correct predictions based on the recovered clean samples. Despite the success of adversarial purification in the computer vision field that incorporates generative models such as energy-based models and diffusion models, using purification as a defense strategy against textual adversarial attacks is rarely explored. In this work, we introduce a novel adversarial purification method that focuses on defending against textual adversarial attacks. With the help of language models, we can inject noise by masking input texts and reconstructing the masked texts based on the masked language models. In this way, we construct an adversarial purification process for textual models against the most widely used word-substitution adversarial attacks. We test our proposed adversarial purification method on several strong adversarial attack methods including Textfooler and BERT-Attack and experimental results indicate that the purification algorithm can successfully defend against strong word-substitution attacks. Copyright © 2022, The Authors. All rights reserved.

关键词： Purification

Constructions of k-uniform states in heterogeneous systems

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Feng, Keqin Jin, Lingfei Xing, Chaoping Yuan, Chen Department of Mathematical Sciences Tsinghua University Beijing China School of Computer Science Shanghai Key Laboratory of Intelligent Information Processing Fudan University Shanghai200433 China The School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai200240 China

A pure quantum state of n parties associated with the Hilbert space Cd1☉ Cd2 ☉ · · · ☉ Cdn is called k-uniform if all the reductions to k-parties are maximally mixed. The n partite system is called homogenous if the local dimension d1 = d2 = · · · = dn, while it is called heterogeneous if the local dimension are not all equal. k-uniform sates play an important role in quantum information theory. There are many progress in characterizing and constructing k-uniform states in homogeneous systems. However, the study of entanglement for heterogeneous systems is much more challenging than that for the homogeneous case. There are very few results known for the k-uniform states in heterogeneous systems for k > 3. We present two general methods to construct k-uniform states in the heterogeneous systems for general k. The first construction is derived from the error correcting codes by establishing a connection between irredundant mixed orthogonal arrays and error correcting codes. We can produce many new k-uniform states such that the local dimension of each subsystem can be a prime power. The second construction is derived from a matrix H meeting the condition that HA×Ā + HĀT×A has full rank for any row index set A of size k. These matrix construction can provide more flexible choices for the local dimensions, i.e., the local dimensions can be any integer (not necessarily prime power) subject to some constraints. Our constructions imply that for any positive integer k, one can construct k-uniform states of a heterogeneous system in many different Hilbert spaces. Copyright © 2023, The Authors. All rights reserved.

关键词： Hilbert spaces

Mitigating Negative Style Transfer in Hybrid Dialogue System

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Shimin Cheng, Qinyuan Li, Linyang Qiu, Xipeng School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

As the functionality of dialogue systems evolves, hybrid dialogue systems that accomplish user-specific goals and participate in open-topic chitchat with users are attracting growing attention. Existing research learns both tasks concurrently utilizing a multi-task fusion technique but ignores the negative transfer phenomenon induced by the unique textual style differences. Therefore, contrastive learning based on the latent variable model is used to decouple the various textual genres in the latent space. We devise supervised and self-supervised positive and negative sample constructions for diverse datasets. In addition, to capitalize on the style information contained in the decoupled latent variables, we employ a style prefix that incorporates latent variables further to control the generation of responses with varying styles. We performed extensive experiments on three dialogue datasets, including a hybrid dialogue dataset and two task-oriented dialogue datasets. The experimental results demonstrate that our method can mitigate the negative style transfer issue and achieves state-of-the-art performance on multiple dialogue datasets. © 2022, CC BY-NC-ND.

关键词： Speech processing

Wind resistance aerial path planning for efficient reconstruction of offshore ship

学校读者我要写书评

暂无评论

International Journal of Digital Earth 2022年第1期15卷 1881-1904页

作者： Tao Liu Ruiqi Shen Zhengling Lei Yuchi Huo Jiansen Zhao Xiaogang Xu College of Transport & Communications Shanghai Maritime UniversityShanghaiPeople’s Republic of China Merchant Marine College Shanghai Maritime UniversityShanghaiPeople’s Republic of China College of Engineering Science and Technology Shanghai Ocean UniversityShanghaiPeople’s Republic of China State Key Lab of CAD&CG Zhejiang UniversityHangzhouPeople’s Republic of China Intelligent Transportation Research Center Zhejiang LabHangzhouPeople’s Republic of China School of Computer and Information Engineering Zhejiang Gongshang UniversityHangzhouPeople’s Republic of China

When the unmanned aerial vehicle(UAV)is applied to three-dimensional(3D)reconstruction of the offshore ship,it faces two problems:the battery capacity limitation of the UAV and the disturbance of the wind in the *** disturbance is generally not considered in the path planning process of the existing UAV 3D reconstruction path planning ***,the planned path is only suitable for no-wind or light-wind *** the 3D reconstruction of ship targets,we propose a UAV path planning method that can satisfy both reconstruction efficiency and wind disturbance resistance ***,the concept of model surface complexity is proposed to generate a more efficient candidate view ***,the Min–Max strategy and a new viewpoint construction method are used to generate the initial ***,combined with the wind field model,a method for generating a stable path against wind disturbance based on the idea of interval optimization is *** results demonstrate that our method can adaptively determine the number of sample points and viewpoints according to ship’s geometric characteristics and further reduce the number of viewpoints without significantly affecting the reconstruction quality;the path planned by our method is also stable against wind disturbance.

关键词： Multi-view 3D reconstruction aerial path planning wind resistance UAV

Late Prompt Tuning: A Late Prompt Could Be Better Than Many Prompts

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liu, Xiangyang Sun, Tianxiang Huang, Xuanjing Qiu, Xipeng School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

Prompt tuning is a parameter-efficient tuning (PETuning) method for utilizing pre-trained models (PTMs) that simply prepends a soft prompt to the input and only optimizes the prompt to adapt PTMs to downstream tasks. Although it is parameter- and deployment-efficient, its performance still lags behind other state-of-the-art PETuning methods. Besides, the training cost of prompt tuning is not significantly reduced due to the back-propagation through the entire model. Through empirical analyses, we shed some light on the lagging performance of prompt tuning and recognize a trade-off between the propagation distance from label signals to the inserted prompt and the influence of the prompt on model outputs. Further, we present Late Prompt Tuning (LPT) that inserts a late prompt into an intermediate layer of the PTM instead of the input layer or all layers. The late prompt is obtained by a neural prompt generator conditioned on the hidden states before the prompt insertion layer and therefore is instance-dependent. Through extensive experimental results across various tasks and PTMs, we show that LPT can achieve competitive performance to full model tuning and other PETuning methods under both full-data and few-shot scenarios while possessing faster training speed and lower memory cost. Copyright © 2022, The Authors. All rights reserved.

关键词： Backpropagation

BERTScore is Unfair: On Social Bias in Language Model-Based Metrics for Text Generation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Sun, Tianxiang He, Junliang Qiu, Xipeng Huang, Xuanjing School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

Automatic evaluation metrics are crucial to the development of generative systems. In recent years, pre-trained language model (PLM) based metrics, such as BERTScore (Zhang et al., 2020), have been commonly adopted in various generation tasks. However, it has been demonstrated that PLMs encode a range of stereotypical societal biases, leading to a concern on the fairness of PLMs as metrics. To that end, this work presents the first systematic study on the social bias in PLM-based metrics. We demonstrate that popular PLM-based metrics exhibit significantly higher social bias than traditional metrics on 6 sensitive attributes, namely race, gender, religion, physical appearance, age, and socioeconomic status. In-depth analysis suggests that choosing paradigms (matching, regression, or generation) of the metric has a greater impact on fairness than choosing PLMs. In addition, we develop debiasing adapters that are injected into PLM layers, mitigating bias in PLM-based metrics while retaining high performance for evaluating text generation. Copyright © 2022, The Authors. All rights reserved.

关键词： Computational linguistics

Multitask Pre-training of Modular Prompt for Chinese Few-Shot Learning

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Sun, Tianxiang He, Zhengfu Zhu, Qin Qiu, Xipeng Huang, Xuanjing School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

Prompt tuning is a parameter-efficient approach to adapting pre-trained language models to downstream tasks. Although prompt tuning has been shown to match the performance of full model tuning when training data is sufficient, it tends to struggle in few-shot learning settings. In this paper, we present Multi-task Pre-trained Modular Prompt (MP2) to boost prompt tuning for few-shot learning. MP2 is a set of combinable prompts pre-trained on 38 Chinese tasks. On downstream tasks, the pre-trained prompts are selectively activated and combined, leading to strong compositional generalization to unseen tasks. To bridge the gap between pre-training and fine-tuning, we formulate upstream and downstream tasks into a unified machine reading comprehension task. Extensive experiments under two learning paradigms, i.e., gradient descent and black-box tuning, show that MP2 significantly outperforms prompt tuning, full model tuning, and prior prompt pre-training methods in few-shot settings. In addition, we demonstrate that MP2 can achieve surprisingly fast and strong adaptation to downstream tasks by merely learning 8 parameters to combine the pre-trained modular prompts. Copyright © 2022, The Authors. All rights reserved.

关键词： Gradient methods

An Embarrassingly Easy but Strong Baseline for Nested Named Entity Recognition

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Yan, Hang Sun, Yu Li, Xiaonan Qiu, Xipeng Shanghai Key Laboratory of Intelligent Information Processing Fudan University China School of Computer Science Fudan University China

Named entity recognition (NER) is the task to detect and classify the entity spans in the text. When entity spans overlap between each other, this problem is named as nested NER. Span-based methods have been widely used to tackle the nested NER. Most of these methods will get a score n × n matrix, where n means the length of sentence, and each entry corresponds to a span. However, previous work ignores spatial relations in the score matrix. In this paper, we propose using Convolutional Neural Network (CNN) to model these spatial relations in the score matrix. Despite being simple, experiments in three commonly used nested NER datasets show that our model surpasses several recently proposed methods with the same pre-trained encoders. Further analysis shows that using CNN can help the model find more nested entities. Besides, we found that different papers used different sentence tokenizations for the three nested NER datasets, which will influence the comparison. Thus, we release a pre-processing script to facilitate future comparison1 © 2022, CC BY-SA.

关键词： Convolutional neural networks

Hierarchical Visual-Textual Graph for Temporal Activity Localization via Language 16th

学校读者我要写书评

暂无评论

Hierarchical Visual-Textual Graph for Temporal Activity Loca...

16th European Conference on computer Vision, ECCV 2020

作者： Chen, Shaoxiang Jiang, Yu-Gang Shanghai Key Lab of Intelligent Information Processing School of Computer Science Fudan University Shanghai China

ISBN: (纸本)9783030585648

Temporal Activity Localization via Language (TALL) in video is a recently proposed challenging vision task, and tackling it requires fine-grained understanding of the video content, however, this is overlooked by most of the existing works. In this paper, we propose a novel TALL method which builds a Hierarchical Visual-Textual Graph to model interactions between the objects and words as well as among the objects to jointly understand the video contents and the language. We also design a convolutional network with cross-channel communication mechanism to further encourage the information passing between the visual and textual modalities. Finally, we propose a loss function that enforces alignment of the visual representation of the localized activity and the sentence representation, so that the model can predict more accurate temporal boundaries. We evaluated our proposed method on two popular benchmark datasets: Charades-STA and ActivityNet Captions, and achieved state-of-the-art performances on both datasets. Code is available at https://***/forwchen/HVTG. © 2020, Springer Nature Switzerland AG.

关键词： Visual languages