Search

Your search keyword '"Liu, Xiaoyun"' showing total 2,867 results
Start Over Author "Liu, Xiaoyun"
2,867 results on '"Liu, Xiaoyun"'

1. Overcoming Growth-Induced Forgetting in Task-Agnostic Continual Learning

Author

Zhao, Yuqing, Saxena, Divya, Cao, Jiannong, Liu, Xiaoyun, and Song, Changlin

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

In continual learning (CL), model growth enhances adaptability over new data, improving knowledge retention for more tasks. However, improper model growth can lead to severe degradation of previously learned knowledge, an issue we name as growth-induced forgetting (GIFt), especially in task-agnostic CL using entire grown model for inference. Existing works, despite adopting model growth and random initialization for better adaptability, often fail to recognize the presence of GIFt caused by improper model growth. This oversight limits comprehensive control of forgetting and hinders full utilization of model growth. We are the first in CL to identify this issue and conduct an in-depth study on root cause of GIFt, where layer expansion stands out among model growth strategies, widening layers without affecting model functionality. Yet, direct adoption of layer expansion presents challenges. It lacks data-driven control and initialization of expanded parameters to balance adaptability and knowledge retention. This paper presents a novel SparseGrow approach to overcome the issue of GIFt while enhancing adaptability over new data. SparseGrow employs data-driven sparse layer expansion to control efficient parameter usage during growth, reducing GIFt from excessive growth and functionality changes. It also combines sparse growth with on-data initialization at training late-stage to create partially 0-valued expansions that fit learned distribution, enhancing retention and adaptability. To further minimize forgetting, freezing is applied by calculating the sparse mask, allowing data-driven preservation of important parameters. Through experiments across datasets with various settings, cases, and task numbers, we demonstrate the necessity of layer expansion and showcase the effectiveness of SparseGrow in overcoming GIFt, highlighting its adaptability and knowledge retention for incremental tasks.

Published
2024

2. NASPrecision: Neural Architecture Search-Driven Multi-Stage Learning for Surface Roughness Prediction in Ultra-Precision Machining

Author

Ruan, Penghui, Saxena, Divya, Cao, Jiannong, Liu, Xiaoyun, Wang, Ruoxin, and Cheung, Chi Fai

Subjects
Computer Science - Computational Engineering, Finance, and Science
Abstract

Accurate surface roughness prediction is critical for ensuring high product quality, especially in areas like manufacturing and aerospace, where the smallest imperfections can compromise performance or safety. However, this is challenging due to complex, non-linear interactions among variables, which is further exacerbated with limited and imbalanced datasets. Existing methods using traditional machine learning algorithms require extensive domain knowledge for feature engineering and substantial human intervention for model selection. To address these issues, we propose NASPrecision, a Neural Architecture Search (NAS)-Driven Multi-Stage Learning Framework. This innovative approach autonomously identifies the most suitable features and models for various surface roughness prediction tasks and significantly enhances the performance by multi-stage learning. Our framework operates in three stages: 1) architecture search stage, employing NAS to automatically identify the most effective model architecture; 2) initial training stage, where we train the neural network for initial predictions; 3) refinement stage, where a subsequent model is appended to refine and capture subtle variations overlooked by the initial training stage. In light of limited and imbalanced datasets, we adopt a generative data augmentation technique to balance and generate new data by learning the underlying data distribution. We conducted experiments on three distinct real-world datasets linked to different machining techniques. Results show improvements in Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Standard Deviation (STD) by 18%, 31%, and 22%, respectively. This establishes it as a robust and general solution for precise surface roughness prediction, potentially boosting production efficiency and product quality in key industries while minimizing domain expertise and human intervention.

Published
2024

6. MGAS: Multi-Granularity Architecture Search for Trade-Off Between Model Effectiveness and Efficiency

Author

Liu, Xiaoyun, Saxena, Divya, Cao, Jiannong, Zhao, Yuqing, and Ruan, Penghui

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Neural architecture search (NAS) has gained significant traction in automating the design of neural networks. To reduce the time cost, differentiable architecture search (DAS) transforms the traditional paradigm of discrete candidate sampling and evaluation into that of differentiable super-net optimization and discretization. However, existing DAS methods fail to trade off between model performance and model size. They either only conduct coarse-grained operation-level search, which results in redundant model parameters, or restrictively explore fine-grained filter-level and weight-level units with pre-defined remaining ratios, suffering from excessive pruning problem. Additionally, these methods compromise search quality to save memory during the search process. To tackle these issues, we introduce multi-granularity architecture search (MGAS), a unified framework which aims to discover both effective and efficient neural networks by comprehensively yet memory-efficiently exploring the multi-granularity search space. Specifically, we improve the existing DAS methods in two aspects. First, we balance the model unit numbers at different granularity levels with adaptive pruning. We learn discretization functions specific to each granularity level to adaptively determine the unit remaining ratio according to the evolving architecture. Second, we reduce the memory consumption without degrading the search quality using multi-stage search. We break down the super-net optimization and discretization into multiple sub-net stages, and perform progressive re-evaluation to allow for re-pruning and regrowing of previous units during subsequent stages, compensating for potential bias. Extensive experiments on CIFAR-10, CIFAR-100 and ImageNet demonstrate that MGAS outperforms other state-of-the-art methods in achieving a better trade-off between model performance and model size.

Published
2023

21. China Alzheimer Report 2024

Author

WANG Gang, QI Jinlei, LIU Xinya, REN Rujing, LIN Shaohui, HU Yisong, LI Haixia, XIE Xinyi, WANG Jintao, LI Jianping, ZHU Yikang, GAO Mengyi, YANG Junjie, WANG Yiran, JING Yurong, GENG Jieli, ZHI Nan, CAO Wenwei, XU Qun, YU Xiaoping, ZHU Yuan, ZHOU Ying, WANG Lin, GAO Chao, LI Binyin, CHEN Shengdi, YUAN Fang, DOU Ronghua, LIU Xiaoyun, LI Xuena, YIN Yafu, CHANG Yan, XU Gang, XIN Jiawei, ZHONG Yanting, LI Chunbo, WANG Ying, ZHOU Maigeng, CHEN Xiaochun, representing the China Alzheimer's Disease Report Writing Group

Subjects
alzheimer's disease, epidemiology, diagnosis, treatment, public health resources, Medicine
Abstract

With the sustained growth of economy and significant changes in social demographics, the issue of elderly-related diseases has increasingly drawn attention particularly. Alzheimer's disease (AD),as a representative disease of neurodegenerative diseases has become a major challenge, affecting the health and quality of life among the elderly population severely. In recent years, the incidence, prevalence, and mortality rate of AD increase in China, imposing substantial economic burdens on families, society, and the entire healthcare system. To proactively address this challenge and respond to the national 'Healthy China Action' initiative, leading experts from Renji Hospital, Shanghai Jiao Tong University School of Medicine,and Chinese Center for Disease Control and Prevention Chronic Non-communicable Disease Control Center, Fudan University School of Public Health, Shanghai Mental Health Center, Ruijin Hospital,Shanghai Jiao Tong University School of Medicine, Fujian Medical University, and other authoritative institutions, have jointly authored the 'China Alzheimer Disease Report 2024'. Building upon previous editions of 2021, 2022, and 2023, this report updates epidemiological data on AD in China, thoroughly analyzes the latest economic burdens of the disease, and comprehensively evaluates the current status of AD diagnosis and treatment services, as well as the allocation of public health resources in our country. The release of the 'China Alzheimer Disease Report 2024' not only reflects China's progress and efforts in AD research and prevention, but also underscores the social heightened concern for elderly health issues. It aims to provide scientific and technical guidance and robust data support for the prevention, diagnosis, and treatment of AD, offering a professional basis for the government and relevant departments to formulate targeted health policies and intervention measures. Furthermore, it serves as a platform for promoting academic exchanges and cooperation in this field domestically and internationally. Through the dissemination and application of this report, we anticipate it will not only serve as a reference for professionals but also enhance public awareness of AD, promote active participation across various sectors of society, and jointly advance the development of elderly health care in China, empowering us towards achieving 'healthy aging'.

Published
2024
Full Text
View/download PDF

22. How to Advance the Improvement of Primary Health Care Model for Rural Populations with Priority Diseases: an Evidence Brief for Person-centered Integrated Health Management Policy

Author

ZHANG Xiaotian, LI Mingyue, LAN Qing, LIU Xiaoyun

Subjects
rural primary health care, exploratory intervention, priority diseases, service utilization, service quality, disease burden, effect evaluation, Medicine
Abstract

Background To better address the problems and challenges facing primary health care services in China, China's National Health Commission and the Gates Foundation have been collaborated to implement primary health care projects in areas such as Shanxi Province, Hubei Province, and part of the rural areas of Henan Province from 2017 to 2022. The overall goal of the project is to explore an effective model for basic health care services and to contribute to health poverty alleviation, which will enhance experience sharing within China and benefit other developing countries. Objective To summarize the experience of implementing person-centered integrated health management project in rural populations with priority diseases, represented by hypertension and diabetes. Methods The collection of information and data for the person-centered integrated health management project encompassed multiple sources, including hospital information system, statistical reporting information, basic public health information system, qualitative interviews, patient surveys, physician surveys and quality surveys of medical records. Descriptive statistical analysis as well as before-and-after comparisons were used as the main evaluation methods. Results The proportion of patients with hypertension and diabetes under standardized management increased steadily, the average hospitalization cost of inpatients decreased, with a gradual increase in healthy living behaviors and a significant improvement medication adherence. In terms of health outcomes, the control rate of hypertension improved significantly but the control rate for blood glucose did not change significantly. Conclusion The service concept and capacity of person-centered integrated health management for priority diseases have been significantly improved. The relevant measures have been transformed into policies to be promoted and implemented in the project areas. Various types of experts are the key factors in promoting the implementation of the project, the sustainability of the measures needs to be maintained.

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results