Search

Your search keyword '"Yang, Wenchuan"' showing total 9 results
Start Over Author "Yang, Wenchuan" Publication Year Range This year
9 results on '"Yang, Wenchuan"'

1. GSpect: Spectral Filtering for Cross-Scale Graph Classification

Author

Zhang, Xiaoyu, Yang, Wenchuan, Feng, Jiawei, Dai, Bitao, Bu, Tianci, and Lu, Xin

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Social and Information Networks
Abstract

Identifying structures in common forms the basis for networked systems design and optimization. However, real structures represented by graphs are often of varying sizes, leading to the low accuracy of traditional graph classification methods. These graphs are called cross-scale graphs. To overcome this limitation, in this study, we propose GSpect, an advanced spectral graph filtering model for cross-scale graph classification tasks. Compared with other methods, we use graph wavelet neural networks for the convolution layer of the model, which aggregates multi-scale messages to generate graph representations. We design a spectral-pooling layer which aggregates nodes to one node to reduce the cross-scale graphs to the same size. We collect and construct the cross-scale benchmark data set, MSG (Multi Scale Graphs). Experiments reveal that, on open data sets, GSpect improves the performance of classification accuracy by 1.62% on average, and for a maximum of 3.33% on PROTEINS. On MSG, GSpect improves the performance of classification accuracy by 15.55% on average. GSpect fills the gap in cross-scale graph classification studies and has potential to provide assistance in application research like diagnosis of brain disease by predicting the brain network's label and developing new drugs with molecular structures learned from their counterparts in other systems.

Published
2024

2. Non-autoregressive Personalized Bundle Generation

Author

Yang, Wenchuan, Yang, Cheng, Li, Jichao, Tan, Yuejin, Lu, Xin, and Shi, Chuan

Subjects
Computer Science - Machine Learning, Computer Science - Information Retrieval
Abstract

The personalized bundle generation problem, which aims to create a preferred bundle for user from numerous candidate items, receives increasing attention in recommendation. However, existing works ignore the order-invariant nature of the bundle and adopt sequential modeling methods as the solution, which might introduce inductive bias and cause a large latency in prediction. To address this problem, we propose to perform the bundle generation via non-autoregressive mechanism and design a novel encoder-decoder framework named BundleNAT, which can effectively output the targeted bundle in one-shot without relying on any inherent order. In detail, instead of learning sequential dependency, we propose to adopt pre-training techniques and graph neural network to fully embed user-based preference and item-based compatibility information, and use a self-attention based encoder to further extract global dependency pattern. We then design a permutation-equivariant decoding architecture that is able to directly output the desired bundle in a one-shot manner. Experiments on three real-world datasets from Youshu and Netease show the proposed BundleNAT significantly outperforms the current state-of-the-art methods in average by up to 35.92%, 10.97% and 23.67% absolute improvements in Precision, Precision+, and Recall, respectively., Comment: Submitted to Information Processing & Management

Published
2024

8. EvilPromptFuzzer: generating inappropriate content based on text-to-image models.

Author

He, Juntao, Dai, Haoran, Sui, Runqi, Yuan, Xuejing, Liu, Dun, Feng, Hao, Liu, Xinyue, Yang, Wenchuan, Cui, Baojiang, and Li, Kedan

Subjects
LANGUAGE models, ANIMAL welfare, RISK of violence, CLOUD computing, RADICALISM
Abstract

Text-to-image (TTI) models provide huge innovation ability for many industries, while the content security triggered by them has also attracted wide attention. Considerable research has focused on content security threats of large language models (LLMs), yet comprehensive studies on the content security of TTI models are notably scarce. This paper introduces a systematic tool, named EvilPromptFuzzer, designed to fuzz evil prompts in TTI models. For 15 kinds of fine-grained risks, EvilPromptFuzzer employs the strong knowledge-mining ability of LLMs to construct seed banks, in which the seeds cover various types of characters, interrelations, actions, objects, expressions, body parts, locations, surroundings, etc. Subsequently, these seeds are fed into the LLMs to build scene-diverse prompts, which can weaken the semantic sensitivity related to the fine-grained risks. Hence, the prompts can bypass the content audit mechanism of the TTI model, and ultimately help to generate images with inappropriate content. For the risks of violence, horrible, disgusting, animal cruelty, religious bias, political symbol, and extremism, the efficiency of EvilPromptFuzzer for generating inappropriate images based on DALL.E 3 are greater than 30%, namely, more than 30 generated images are malicious among 100 prompts. Specifically, the efficiency of horrible, disgusting, political symbols, and extremism up to 58%, 64%, 71%, and 50%, respectively. Additionally, we analyzed the vulnerability of existing popular content audit platforms, including Amazon, Google, Azure, and Baidu. Even the most effective Google SafeSearch cloud platform identifies only 33.85% of malicious images across three distinct categories. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

9. Nonrepresentativeness of Human Mobility Data and its Impact on Modeling Dynamics of the COVID-19 Pandemic: Systematic Evaluation.

Author

Liu C, Holme P, Lehmann S, Yang W, and Lu X

Abstract

Background: In recent years, a range of novel smartphone-derived data streams about human mobility have become available on a near-real-time basis. These data have been used, for example, to perform traffic forecasting and epidemic modeling. During the COVID-19 pandemic in particular, human travel behavior has been considered a key component of epidemiological modeling to provide more reliable estimates about the volumes of the pandemic's importation and transmission routes, or to identify hot spots. However, nearly universally in the literature, the representativeness of these data, how they relate to the underlying real-world human mobility, has been overlooked. This disconnect between data and reality is especially relevant in the case of socially disadvantaged minorities., Objective: The objective of this study is to illustrate the nonrepresentativeness of data on human mobility and the impact of this nonrepresentativeness on modeling dynamics of the epidemic. This study systematically evaluates how real-world travel flows differ from census-based estimations, especially in the case of socially disadvantaged minorities, such as older adults and women, and further measures biases introduced by this difference in epidemiological studies., Methods: To understand the demographic composition of population movements, a nationwide mobility data set from 318 million mobile phone users in China from January 1 to February 29, 2020, was curated. Specifically, we quantified the disparity in the population composition between actual migrations and resident composition according to census data, and shows how this nonrepresentativeness impacts epidemiological modeling by constructing an age-structured SEIR (Susceptible-Exposed-Infected- Recovered) model of COVID-19 transmission., Results: We found a significant difference in the demographic composition between those who travel and the overall population. In the population flows, 59% (n=20,067,526) of travelers are young and 36% (n=12,210,565) of them are middle-aged (P<.001), which is completely different from the overall adult population composition of China (where 36% of individuals are young and 40% of them are middle-aged). This difference would introduce a striking bias in epidemiological studies: the estimation of maximum daily infections differs nearly 3 times, and the peak time has a large gap of 46 days., Conclusions: The difference between actual migrations and resident composition strongly impacts outcomes of epidemiological forecasts, which typically assume that flows represent underlying demographics. Our findings imply that it is necessary to measure and quantify the inherent biases related to nonrepresentativeness for accurate epidemiological surveillance and forecasting., (©Chuchu Liu, Petter Holme, Sune Lehmann, Wenchuan Yang, Xin Lu. Originally published in JMIR Formative Research (https://formative.jmir.org), 28.06.2024.)

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results