AI Article Synopsis
Article Abstract
Random feature (RF) has been widely used for node consistency in decentralized kernel ridge regression (KRR). Currently, the consistency is guaranteed by imposing constraints on coefficients of features, necessitating that the RFs on different nodes are identical. However, in many applications, data on different nodes vary significantly on the number or distribution, which calls for adaptive and data-dependent methods that generate different RFs. To tackle the essential difficulty, we propose a new decentralized KRR algorithm that pursues consensus on decision functions, which allows great flexibility and well adapts data on nodes. The convergence is rigorously given, and the effectiveness is numerically verified: by capturing the characteristics of the data on each node, while maintaining the same communication costs as other methods, we achieved an average regression accuracy improvement of 25.5% across six real-world datasets.
Download full-text PDF |
Source |
|---|---|
| http://dx.doi.org/10.1109/TNNLS.2024.3414325 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Analysis on the spatiotemporal pattern and driving force of the spatial deviation index of grain and economy in counties in China.
PLoS One
December 2024
Sichuan Agricultural University, Ya'an, Sichuan, China.
Studying the spatial relationship and driving forces between grain production and economic development in China can assist in the coordinated development of economic growth and grain production in both China and other developing countries. Based on panel data from 2000 to 2019 covering 2018 county-level units in China, this study comprehensively investigated the spatial distribution, spatial differences, dynamic evolution of distribution, and driving factors of China's county-level spatial deviation index of grain and economy (SDIGE) using methods such as the standard deviation ellipse method, the three-stage nested decomposition of Theil index, kernel density estimation, and geographically weighted regression (GWR) model. The results show that (1) from 2000 to 2019, China's SDIGE showed a development trend of "up-down-up," and the highest SDIGE was in the northeast region, the lowest in the east region, and the spatial pattern of "high in the northeast-low in the east coast" was increasingly prominent.
View Article and Find Full Text PDFKernel-Based Decentralized Policy Evaluation for Reinforcement Learning.
IEEE Trans Neural Netw Learn Syst
September 2024
Article Synopsis
- The research explores how multiple agents can work together in reinforcement learning to evaluate state-value functions while keeping their rewards private and using sampled state transitions.
- The study employs a regression-based iteration technique with gradient descent in a mathematical framework called reproducing kernel Hilbert space, and uses Nyström approximation to simplify calculations.
- The paper establishes new statistical error bounds for estimating value functions in decentralized settings and compares their method to previous approaches, specifically kernel temporal difference methods.
Decentralized Kernel Ridge Regression Based on Data-Dependent Random Feature.
IEEE Trans Neural Netw Learn Syst
July 2024
Random feature (RF) has been widely used for node consistency in decentralized kernel ridge regression (KRR). Currently, the consistency is guaranteed by imposing constraints on coefficients of features, necessitating that the RFs on different nodes are identical. However, in many applications, data on different nodes vary significantly on the number or distribution, which calls for adaptive and data-dependent methods that generate different RFs.
View Article and Find Full Text PDFDynamic evolution and spatial difference of public health service supply in economically developed provinces of China: typical evidence from Guangdong Province.
BMC Health Serv Res
January 2024
School of Health Management, Southern Medical University, Guangzhou510515, China.
Objective: The outbreak of the COVID-19 pandemic has drawn attention from all sectors of society to the level of public health services. This study aims to investigate the level of public health service supply in the four major regions of Guangdong Province, providing a basis for optimizing health resource allocation.
Methods: This article uses the entropy method and panel data of 21 prefecture-level cities in Guangdong Province from 2005 to 2021 to construct the evaluation index system of public health service supply and calculate its supply index.
Spatial distribution pattern and driving mechanism of tourist attractions in Gansu Province based on POI data.
PLoS One
November 2023
Tourism College, Northwest Normal University, Lanzhou, Gansu, China.
The article utilizes POI (Point of Interest) data of tourist attractions in Gansu Province in 2021, adopts Moran's I and kernel density analysis to study the spatial distribution pattern of tourist attractions in Gansu Province, and uses spatial autoregressive modeling to explore the driving mechanism affecting their spatial distribution pattern. The results show that: (1) Gansu Province has a large number and rich types of tourist attractions, and there are differences in the number of different types of tourist attractions; (2) The spatial distribution pattern of different types of tourist attractions in different cities and towns shows the phenomenon of both agglomeration and dispersion, with a higher degree of agglomeration in the central and northwestern regions of the province and a lower degree of agglomeration in the southwestern and southeastern corners; (3) The overall spatial distribution pattern of tourist attractions shows the distribution characteristics of multi-core decentralized distribution, forming 8 core aggregation areas in the southeast of the province; (4) The article analyzes the driving mechanism of the spatial distribution pattern of tourist attractions in Gansu Province using the buffer zone and OLS models, and the results show that the natural environment, transportation location, national policies and socio-economics all have a positive impact on the distribution of tourist attractions.
View Article and Find Full Text PDFWant AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!