2个月没关注，今天想访问一下，发现打不开，不知是否是一个牺牲品？
可怜的网赚之家
网赚之家 打不开
2010年06月24日最近拍适合初学者看的双节棍教材
2010年05月2日KDD 2010 – Accepted Papers
2010年05月2日Research Full Presentations
# A Hierarchical Information Theoretic Technique for the Discovery of Non Linear Alternative Clusterings
Xuan Hong Dang*, The University of Melbourne; James Bailey, The University of Melbourne
# A Scalable Two-Stage Approach for a Class of Dimensionality Reduction Techniques
Liang Sun*, Arizona State University; Betul Ceran, Arizona State University; Jieping Ye, Arizona State University
# A Statistical Model for Popular Event Tracking in Social Communities
Xide Lin*, UIUC; Bo Zhao, U of Illinois,Urbana Champaign; Qiaozhu Mei, Univ. of Michigan; Jiawei Han,
# An Efficient Algorithm for a Class of Fused Lasso Problems
Jun Liu*, ASU; Lei Yuan, ; Jieping Ye, Arizona State University
# An efficient causal discovery algorithm for linear models
Zhenxing Wang*, The Chinese University of Hong; Laiwan Chan, The Chinese University of Hong Kong
# An Energy-Efficient Mobile Recommender System
Yong Ge*, Rutgers University; Hui Xiong, Rutgers University; Alexander Tuzhilin, Stern School of Business, New York University; Keli Xiao, Rutgers University; Marco Gruteser, Rutgers University
# Balanced Allocation with Succinct Representation
Saeed Alaei, University of Maryland; Ravi Kumar*, Yahoo; Azaraksh Malekian, malekian@cs.umd.edu; Erik Vee, Yahoo! Research
# Class-Specific Error Bounds for Ensemble Classifiers
Ryan Prenger*, Lawrence Livermore National La; Tracy Lemmond, Lawrence Livermore National Laboratory; Barry Chen, Lawrence Livermore National Laboratory; Kush Varshney, Massachusetts Institute of Technology; William Hanley, Lawrence Livermore National Laboratory
# Clustering by Synchronization
Christian Böhm*, University of Munich; Claudia Plant, Technische Universität München; Junming Shao, University of Munich; Qinli Yang, University of Edinburgh
# Collusion-Resistant Privacy-Preserving Data Mining
Bin Yang*, The University of Tokyo; Hiroshi Nakagawa, ; issei Sato, ; Jun Sakuma, University of Tsukuba
# Combined Regression and Ranking
D. Sculley*, Google, Inc
# Combining Predictions for Accurate Recommender Systems
Michael Jahrer*, Commendo research & consulting; Andreas Töscher, Commendo research & consulting; Robert Legenstein, Graz University of Technology
# Community Outliers and their Efficient Detection in Information Networks
Jing Gao*, UIUC; Feng Liang, UIUC; Wei Fan, IBM T.J.Watson; Chi Wang, UIUC; Yizhou Sun, University of Illinois at Urbana Champaign; Jiawei Han, UIUC
# Compressed Fisher Linear Discriminant Analysis: Classification of Randomly Projected Data
Robert Durrant*, University of Birmingham; Ata Kaban, University of Birmingham
# Connecting the Dots Between News Articles
Dafna Shahaf*, CMU; Carlos Guestrin, CMU
# Data Mining with Differential Privacy
Arik Friedman*, Technion; Assaf Schuster, Technion
# Designing efficient cascaded classifiers: Tradeoff between accuracy and cost
Vikas Raykar*, Siemens Healthcare; Balaji Krishnapuram, Siemens Healthcare; Shipeng Yu, Siemens Healthcare
# Discovering frequent patterns in sensitive data
Raghav Bhaskar, Microsoft Research; Srivatsan Laxman*, Microsoft Research; Adam Smith, Pennsylvania State University; Abhradeep Thakurta, Pennsylvania State University
# Discovering Significant Relaxed Order-Preserving Submatrices
Qiong FANG*, HKUST; Wilfred Ng, Hong Kong UST; Jianlin Feng, Sun Yat-sen University
# Discriminative Topic Modeling based on Manifold Learning
Seungil Huh*, Carnegie Mellon University; Stephen Fienberg,
# Document Clustering via Dirichlet Process Mixture Model with Feature Selection
Guan Yu, ; Ruizhang Huang*, The Hong Kong Polytechnic Univ; Zhaojun Wang,
# DUST: A Generalized Notion of Similarity between Uncertain Time Series
Smruti Sarangi, IBM Research – India; Karin Murthy*, IBM Research – India
# Estimating Rates of Rare Events with Multiple Hierarchies through Scalable Log-linear Models
Deepak Agarwal*, ; Nagaraj Kota, ; Rahul Agrawal, ; Rajiv Khanna,
# Evolutionary Hierarchical Dirichlet Processes for Multiple Correlated Time-varying Corpora
Jianwen Zhang*, Tsinghua University; Yangqiu Song, ; Changshui Zhang, Tsinghua University; Shixia Liu,
# Extracting Temporal Signatures for Comprehending Systems Biology Models
Naren Sundaravaradan, Virginia Tech; K. S. M. Tozammel Hossain, Virginia Tech; Vandana Sreedharan, Virginia Tech; John Paul Vergara, Ateneo de Manila University; Lenwood Heath, Virginia Tech; Douglas Slotta, NIH/NCBI; Naren Ramakrishnan*, Virginia Tech
# Fast Euclidean Minimum Spanning Tree: Algorithm, Analysis, Applications
William March*, Georgia Institute of Technolog; Parikshit Ram, Georgia Institute of Technology; Alexander Gray, Georgia Institute of Technology
# Fast Nearest Neighbor Search in Disk-resident Graphs
Purnamrita Sarkar*, CMU; Andrew Moore, Google
# Fast Online Learning through Effective Offline Initialization for Time-Sensitive Recommendation
Bee-Chung Chen*, Yahoo! Research; Deepak Agarwal, ; Pradheep Elango, Yahoo! Labs
# Fast Query Execution for Retrieval Models based on Path Constraint Random Walks
Ni Lao*, Carnegie Mellon University; William Cohen, Carnegie Mellon University
# Flexible Constrained Spectral Clustering
Xiang Wang*, UC Davis; Ian Davidson, UC Davis
# Frequent Regular Itemset Mining
Salvatore Ruggieri*, Università di Pisa
# GLS-SOD: A Generalized Local Statistical Approach for Spatial Outlier Detection
Feng Chen*, Virginia Tech; Chang-Tien Lu, Virginia Tech
# Grafting-Light: Fast, Incremental Feature Selection and Structure Learning of Markov Random Fields
Jun Zhu*, Carnegie Mellon University; Ni Lao, Carnegie Mellon University; Eric Xing, Carnegie Mellon Univresity
# Growing a tree in the forest: constructing folksonomies by integrating structured metadata
Anon Plangprasopchok*, Information Sciences Institute; Kristina Lerman, USC; Lise Getoor, University of Maryland, College Park
# Inferring Networks of Diffusion and Influence
Manuel Gomez Rodriguez*, Stanford University; Jure Leskovec, Stanford University; Andreas Krause, California Institute of Technology
# k-Support Anonymity based on Pseudo Taxonomy for Outsourcing of Frequent Itemset Mining
Chih-Hua Tai*, Ntu; Philip Yu, University of Illinois at Chicago; Ming-Syan Chen,
# Large Linear Classification When Data Cannot Fit In Memory
Hsiang-Fu Yu*, National Taiwan University; Cho-Jui Hsieh , ; Kai-Wei Chang, ; Chih-Jen Lin, National Taiwan University
# Learning Incoherent Sparse and Low-Rank Patterns from Multiple Tasks
Jianhui Chen*, Arizona State University; Ji Liu, Arizona State University; Jieping Ye, Arizona State University
# Learning to Combine Discriminative Classifiers
Chi-Hoon Lee*, Yahoo! Labs
# Learning with Cost Intervals
Xu-Ying Liu*, Nanjing University; Zhi-Hua Zhou, Nanjing University
# Mass Estimation and Its Applications
Kai Ming Ting*, Monash University; Guang-Tong Zhou, Shandong University; Fei Tony LIU, Monash University; James Tan, Monash University
# Mining Advisor-Advisee Relationships from Research Publication Networks
Chi Wang*, UIUC; Jiawei Han, ; Yuntao Jia, ; Jie Tang, Tsinghua; Duo Zhang, UIUC; Yintao Yu, UIUC; Jingyi Guo,
# Mining Positive and Negative Patterns for Relevance Feature Discovery
Yuefeng Li*, Queensland University of Techn; Abdulmohsen Algarni, ; Ning Zhong, Maebashi Institute of Technology, Japan
# Mining Program Workflow from Interleaved Traces
Jian-Guang LOU*, Microsoft Research Asia; Qiang FU, Microsoft Research Asia; Shengqi YANG, Beijing Univ. of Posts and Telecom; Jiang LI, Microsoft Research Asia; Bin WU, Beijing Univ. of Posts and Telecom
# Mining Top-K Frequent Items in a Data Stream with Flexible Sliding Windows
Hoang Thanh Lam*, TU Eindhoven; Toon Calders, technische Universiteit Eindhoven
# Mining Uncertain Data with Probabilistic Guarantees
Liwen Sun*, University of Hong Kong; Reynold Cheng, University of Hong Kong; David Cheung, University of Hong Kong; Jiefeng Cheng,
# Modeling Relational Events via Latent Classes
Christopher DuBois*, UC Irvine; Padhraic Smyth,
# Multi-Label Learning by Exploiting Label Dependency
Min-Ling Zhang*, Hohai University; Kun Zhang, MPI for Biological Cybernetics
# Multi-Task Learning for Boosting with Application to Web Search Ranking
Olivier Chapelle*, Yahoo! Research; Srinivas Vadrevu, Yahoo! Labd; Kilian Weinberger, Washington University in St. Louis; Pannagadatta Shivaswamy, Columbia University; Ya Zhang, Shanghai Jiaotong University; Belle Tseng, Yahoo! Labs
# Negative correlations in collaboration: concepts and algorithms
Jinyan Li*, Nanyang Technological University, Singapore; Qian Liu, NTU; Tao Zeng, NTU
# Neighbor Query Friendly Compression of Social Networks
Hossein Maserrat*, Simon Fraser University; Jian Pei, SFU
# Nonnegative Shared Subspace Learning and Its Application to Social Media Retrieval
Sunil Gupta*, Curtin University; Dinh Phung, Curtin University; Brett Adams, Curtin University; Truyen Tran, Curtin University; Svetha Venkatesh, Curtin University
# On the Quality of Inferring Interests From Social Neighbors
Zhen Wen*, IBM T.J. Watson Research; Ching-Yung Lin, IBM T.J. Watson Research Center
# Online Discovery and Maintenance of Time Series Motifs
Abdullah Mueen*, UC Riverside; Eamonn Keogh, UC Riverside
# Online Multiscale Dynamic Topic Models
Tomoharu Iwata*, ; Takeshi Yamada, NTT; Yasushi Sakurai, NTT; Naonori Ueda, NTT
# Oracle Classification – Learning What Really Matters
Ulf Johansson*, University of Boras; Cecilia Sönströd, ; Tuve Löfström,
# Privacy-Preserving Outsourcing Support Vector Machines with Random Transformation
Ming-Syan Chen*, ; Keng-Pei Lin, National Taiwan University
# Redefining Class Definitions using Constraint-Based Clustering
Dan Preston*, Tufts University; Carla Brodley, Tufts University; Roni Khardon, Tufts University; Damien Sulla-Menashe, Boston University; Mark Friedl, Boston University
# Scalable Influence Maximization for Prevalent Viral Marketing in Large-Scale Social Networks
Wei Chen, ; Chi Wang, UIUC; Yajun Wang*,
# Scalable Similarity Search with Optimized Kernel Hashing
Junfeng He*, Columbia University; Wei Liu, Columbia University; Shih-Fu Chang, Columbia University
# Semi-Supervised and Sparse Metric Learning Using Alternating Direction Optimization
Wei Liu*, CUHK; Shiqian Ma, ; Dacheng Tao, Nanyang Technological University; Jianzhuang Liu,
# Semi-supervised Feature Selection for Graph Classification
Xiangnan Kong, University of Illinois; Philip Yu*, University of Illinois at Chicago
# Suggesting Friends Using the Implicit Social Graph
Maayan Roth*, Google; Assaf Ben-David, Google; David Deutscher, Google, Inc; Ilan Horn, Google, Inc; Aril Leichtberg, Google; Naty Leiser, Google; Ron Merom, Google; Yossi Mattias, Google, Inc
# The community-search problem and how to plan a successful cocktail party
Mauro Sozio, Max-Planck-Institut fur Informatik; Aristides Gionis*, Yahoo! Research Barcelona
# The new Iris Data: Modular Data Generators
Iris Adae*, Universitaet Konstanz; Michael Berthold, University of Konstanz
# The Topic-Perspective Model for Social Tagging Systems
Caimei Lu*, Drexel University; Xiaohua Hu, Drexel University; Xin Chen, Drexel University; Jung-ran Park, Drexel University
# Topic Dynamics: an alternative model of `Bursts’ in Streams of Topics
Dan He*, UCLA; Douglass Parker, UCLA Computer Science Dept
# Topic Models with Power-Law Using Pitman-Yor Process
Issei Sato*, University of Tokyo; Hiroshi Nakagawa, University of Tokyo
# Training and Testing of Recommender Systems on Data Missing Not at Random
Harald Steck*, Bell Labs, Alcatel-Lucent
# Trust Network Inference for Online Rating Data Using Generative Models
Freddy Chong Tat Chua*, Singapore Management Universit; Ee-Peng Lim, Singapore Management University
# Unifying Dependent Clustering and Disparate Clustering for Non-homogeneous Data
M. Shahriar Hossain, Virginia Tech; Satish Tadepalli, Virginia Tech; Layne Watson, Virginia Tech; Ian Davidson, UC Davis; Richard Helm, Virginia Tech; Naren Ramakrishnan*, Virginia Tech
# Unsupervised Feature Selection for Multi-Cluster Data
Deng Cai*, Zhejiang University; Chiyuan Zhang, Zhejiang University; Xiaofei He, Zhejiang University
# Unsupervised Transfer Learning: Application to Text Categorization
Tianbao Yang*, Michigan State University; Rong Jin, Michigan State University; Anil Jain, Michigan State University; Yang Zhou, Michigan State University; Wei Tong, Michigan State University
# UP-Growth: An Efficient Algorithm for High Utility Itemsets Mining
Vincent Tseng*, National Cheng Kung University; Cheng Wei Wu, National Cheng Kung University; Bai-En Shie, National Cheng Kung University; Philip Yu, University of Illinois at Chicago
# User Browsing Models: Relevance versus Examination
Ramakrishnan Srikant*, Google Research; Sugato Basu, Google Research; Ni Wang, ; Daryl Pregibon, “Google, USA”
# Versatile Publishing for Privacy Preservation
Xin Jin*, George Washington University; Mingyang Zhang, George Washington University; Nan Zhang, George Washington University; Gautum Das, UT Arlington
# Why label when you can search? Strategies for applying human resources to build classification models under extreme class imbalance.
Josh Attenberg*, NYU Polytechnic Institute; Foster Provost, NYU
Research Short Presentations
# A POWER Framework for Multi-Class Membership in Bayesian Mixture Models
Manas Somaiya*, University of Florida; Christopher Jermaine, Rice University; Sanjay Ranka, University of Florida
# A Probabilistic Model for Personalized Tag Prediction
Dawei Yin*, Lehigh University; Zhenzhen Xue, Lehigh University; Liangjie Hong, Lehigh University; Brian Davison, Lehigh University
# A Unified Algorithmic Framework for Multi-Dimensional Scaling
Arvind Agarwal, University of Utah; Jeff Phillips*, University of Utah; Suresh Venkatasubramanian, University of Utah
# BioSnowball: Automated Population of Wikis
Xiaojiang Liu, ; Zaiqing Nie*, Microsoft; Nenghai Yu, ; Ji-Rong Wen, Microsoft Research Asia
# Boosting with Structure Information in the Functional Space: an Application to Graph Classification
Hongliang Fei*, University of Kansas; Jun Huan, University of Kansas
# Cold Start Link Prediction
Vincent Leroy, IRISA; Berkant Cambazoglu, Yahoo! Research; Francesco Bonchi*, Yahoo! Research
# Community-based Greedy Algorithm for Mining Top-K Influential Nodes in Mobile Social Networks
Yu Wang*, PKU; Gao Cong, Nanyang Techonological University; Guojie Song, Peking University; Kunqing Xie, Peking University
# Direct Mining of Discriminative Patterns for Classifying Uncertain Data
Chuancong Gao*, Tsinghua University; Jianyong Wang, Tsinghua University
# Discovering Probabilistic Frequent Subgraphs over Uncertain Graph Databases
Zhaonian Zou*, Harbin Institute of Technology; Jianzhong Li, Harbin Institute of Technology; Hong Gao, Harbin Institute of Technology
# DivRank: the Interplay of Prestige and Diversity in Information Networks
Qiaozhu Mei*, Univ. of Michigan; Jian Guo, University of Michigan; Dragomir Radev, University of Michigan
# Dynamics of Conversations
Ravi Kumar, Yahoo; Mohammad Mahdian, Yahoo! Research; Mary McGlohon*,
# Ensemble Pruning via Individual Contribution Ordering
Zhenyu Lu*, University of Vermont; Xindong Wu, University of Vermont; Josh Bongard, University of Vermont
# Feature Selection for Support Vector Regression Using Probabilistic Prediction
Chong-Jin Ong*, National University Singapore; Jianbo Yang, National University Singapore
# Finding Effectors in Social Networks
Theodoros Lappas*, UCR; Heikki Mannila, ; Evimaria Terzi, Boston University; Dimitrios Gunopulos, UoA
# Generative Models for Ticket Resolution in Expert Networks
Gengxin Miao*, UC at Santa Barbara; Louise Moser, UC at Santa Barbara; Xifeng Yan, University of California at Santa Barbara; Shu Tao, IBM T. J. Watson; Yi Chen, Arizona State Univ.; Nikos Anerousis, IBM T. J. Watson
# Latent Aspect Rating Analysis on Review Text Data: A Rating Regression Approach
Hongning Wang*, University of Illinois; Yue Lu, University of Illinois; ChengXiang Zhai, UIUC
# New Perspectives and Methods in Link Prediction
Ryan Lichtenwalter*, The University of Notre Dame; Jake Lussier, The University of Notre Dame; Nitesh Chawla, The University of Notre Dame
# Parallel SimRank Computation on Large Graphs with Iterative Aggregation
Guoming He, Renmin University of China; Haijun Feng, ; Cuiping Li*, Renmin University of China; Hong Chen,
# Probably the Best Itemsets
Nikolaj Tatti*, University of Antwerp
# Semantic Relation Extraction With Kernels Over Typed Dependency Trees
Frank Reichartz*, Fraunhofer IAIS; hannes Korte, Fraunhofer IAIS; Gerd Paass,
# Social Action Tracking via Noise Tolerant Time-varying Factor Graphs
Chenhao Tan, Tsinghua University; Jie Tang*, Tsinghua; Jimeng Sun, IBM; Quan Lin, Huazhong University of Science and Technology; Fengjiao Wang, BeiJing University of Aeronautics & Astronautics
# Temporal Recommendation on Graphs via Long- and Short-term Preference Fusion
Liang Xiang, Institute of Automation, Chinese Academy of Sciences; Quan Yuan*, IBM Research – China; Shiwan Zhao, IBM Research – China; Li Chen, Department of Computer Science, Hong Kong Baptist University; Xiatian Zhang, IBM Research – China; Jimeng Sun, IBM
# Towards Mobility-based Clustering
Siyuan Liu*, HKUST; Yunhuai Liu, ; Lionel Ni, ; Jianping Fan, ; Minglu Li,
# Transfer Metric Learning by Learning Task Relationships
Yu Zhang*, HKUST; Dit-Yan Yeung, Hong Kong University of Science and Technology
科学中药计算器
2010年05月1日使用中成药来匹配经方的计算方式：
中医天下
2010年04月28日在线全文阅读：http://www.3320.net/blib/c/read/18/8677/ra8677.htm#3094
or http://read.shanwen.com/14/14781/1052145.html
真正写中医小说写得最好的，还是西川的神针记，看完让人欲罢不能。很是期待其后续作品，可惜近一年悄无声息。
IPv6现状
2010年04月28日http://www.deepspace6.net/docs/ipv6_status_page_apps.html 给出了目前支持IPv6协议的各个层次的应用程序。
最后更新时间是2010年3月9号
[Tccc] Cfp: CSCN2010 – Aug 15 2010 – Beijing, China
2010年04月22日The First Workshop on Compressive Sensing for Communications and Networking
Aug 15, 2010, Beijing, China
http://students.uta.edu/jx/jxl5466/cscn-wasa10.htm
Call for Papers
———————————————-
Compressive Sensing (CS), also known as compressive sampling, is a novel
sensing/sampling paradigm
that goes against the common wisdom in data acquisition. CS theory asserts
that one can recover certain
signals and images from far fewer samples or measurements than traditional
methods. The crucial
observation is that one can design efficient sensing or sampling protocols
that capture the useful
information content embedded in a sparse signal and condense it into a small
amount of data. The goal of
this workshop is to disseminate the most recent results in the development
of CS in communications and
networks. Researchers and practitioners working in this area are expected to
take this opportunity to discuss
and express their views on the current trends, challenges, and state of the
art solutions addressing various
issues in CS for communications and networks. Review papers on CS are also
welcome. Topics to be
covered in this workshop include but are not limited to:
* CS for sensor networks
* CS for wireless communications
* CS for channel equalization
* CS for joint source and channel coding
* CS for MIMO systems
* CS for wireless mesh networks
* CS for cellular systems
* CS for channel modeling
* Information theory of CS
* CS for network traffic
* CS for multimedia traffic
* CS for ultra-wideband systems
* CS for remote sensing applications
* CS for nonadaptive signal compression or streaming dataset reduction
* CS for Internet
* CS in multimedia communications
* CS for wired communications and networks
* CS for spread spectrum
* CS for interference cancelation
* CS for co-existence of wireless systems
* CS methods that are tolerant to noise, signal nonsparsity, or measurement
nonlinearity in
communications
* hardware implementation of CS systems
* Other applications
Submission:
Sunday, May 16, 2010
Notification:
Tuesday, June 1, 2010
Camera ready:
Tuesday, June 15, 2010
General Chair:
Prof. Jing Liang, University of Texas at Arlington, USA
E-mail: jliang@wcn.uta.edu
TPC Chairs:
Prof. Dechang Chen, Uniformed Services University of the Health Sciences,
USA
Email: dchen@usuhs.mil
Prof. Qilian Liang, University of Texas at Arlington, USA
E-mail: liang@uta.edu
Program Committee:
Dr. Xiuzhen Cheng, George Washington University
Dr. Ting Jiang, Beijing University of Posts and Telecommunications, China
Dr. Qingchun Ren, Microsoft, Seattle, USA
Dr. Sherwood W. Samn, Air Force Research Laboratory/RHX, San Antonio, TX
78235, USA
Dr. Lingming Wang, iBiquity Digital Corporation, Basking Ridge, NJ 07920 USA
Dr. Xinsheng Xia, Tellabs Inc, New Jersey, USA
Dr. Liang Zhao, Airvana Inc, Chelmsford, MA
Dr. Zheng Zhou, Beijing University of Posts and Telecommunications, China
Belief propagation
2010年04月22日http://blog.sina.com.cn/s/blog_60a751620100eiq8.html
(zz from Dahua’s blog)
Belief propagation是machine learning的泰斗J. Pearl的最重要的贡献。对于统计学来说，它最重要的意义就是在于提出了一种很有效的求解条件边缘概率(conditional marginal probability)的方法。说的有点晦涩了，其实所谓求解条件边缘概率，通俗地说，就是已知某些条件的情况下，推导另外某些事件发生的概率。
如果涉及的因素只有那么几个，一个学过概率论的大学生就可以使用简单的概率公式计算出来。可是，在现实世界中有成千上万的因素，它们相互联系，如果按照传统方法，就要对数以千计的变量进行积分。考虑到运算量对于变量个数以指数增长，因此这么做实际上根本没法算的。虽然，后来人们提出了蒙特卡罗(Monte Carlo)积分，但是对于拥有数以千计变量的复杂系统，仍然可以说是computationally prohibitive。
这个困难一直阻碍着统计推断方法在大规模系统中的应用。Belief propagation出来之后，情况才发生了转变。J. Pearl在他的书中分析说，人们在头脑中经常进行各种各样的推断，可是人在头脑里面发生了什么事情呢：穷举所有未知变量的可能状态进行积分(Traditional method)？还是随即产生各种状态求均值(Monte Carlo Integral)，看来都不make sense。J. Pearl认为，虽然影响世界的因素繁多，但是每个因素实际上只与少数几个因素相关，这就构成了一个推断网络。在machine learning里面，这样的网络有两种：Bayesian Network，反映的是因果推断关系（就是说，相互联系的因素中，其中一个是因，另外一个是果），以及Markov Network，反映的是相互影响的关系（两个因素互为因果，其变化相互影响）。根据这种建模方式，J.Pearl提出把inference局部化和分布化，把全局的积分变成局部的消息传递。网络中的每个节点通过和邻近节点交换信息对自身的概率状况进行评估。通过这种方式，使得计算量从指数增长变成近似的线性增长，从而使得统计推断能在复杂系统中被应用。
数学上可以证明，对于有向无环的Bayesian Network，可以证明，通过BP得到的解和严格的积分计算得到的结果是一致的。这时的BP只是利用因素联系的局部性来简化计算，并把计算过程分散到各个节点。对于无向而且到处是环的markov network，J.Pearl指出，这种传播过程是可能导致不稳定的。某些消息可能在环状的传播过程中无限加强，从而导致整个系统发散或者偏离。但是实际经验表明，对于大部分问题，BP在带环的系统中依然工作良好。很多人对这个现象进行了研究，对于某些特例给出了初步的解释，但是关于Loopy BP的稳定性和收敛性问题，离理论上的最终解决，还有很长的路要走。
在computer vision领域，MIT的著名教授W.T. Freeman是BP方法的积极倡导者，他大量使用markov random field和belief propagation对图像进行建模，在很多应用领域取得了不错的结果。
其实关于Local propagation的方法论，现在有超出了belief propagation的范畴。某些新的方法的优化流程也体现了类似的特征。我在我自己最近一项关于应用信息论进行监督学习的工作中发现，如果使用信息论最大化引导分类过程，事实上其优化流程就体现了样本间的局部交互和传播。不仅仅是机器学习，物理学也同样如此，比如当一个粒子发生了运动，其效应也是不断通过与邻近粒子的相互作用向外扩散，从而形成波动过程。一直以来，machine learning的formulation都是对问题从宏观着眼，可是最近一些工作都在暗示着微观作用过程对于宏观状态形成的重要意义。因此，一些新的研究开始偏向对局部结构和微观作用的考察，得到了一些有趣的发现。Manifold Learning中的一些重要方法，比如Local Linearly Embedding就是其中重要的代表。
被忽视的大肠经——值得敬重的人体血液清道夫zz
2010年04月21日http://51yam.com/viewthread.php?tid=6397&extra=&page=1
一个团体总有被忽视的成员，他们总是在那里默默无闻地工作，很少有出头露面的机会。看起来他们似乎无足轻重，位卑言轻，但他们的作用，却是不可或缺，有时甚至是无可替代。大肠经就是这样一个无名英雄，好像没有什么广大而显赫的功效，但有些特殊的疾病，真得它亲自出马才行。
皮肤病可以说是最让人心烦意乱的疾病了，荨麻疹、神经性皮炎、日光性皮炎、牛皮癣、疥疮、丹毒、疖肿、皮肤瘙痒症……都让人痛苦不堪。在百治无效之际，取大肠经刮痧，通常都会得到不同程度的缓解。 3 |8 `4 I6 `) S; O$ b
大肠经为多气多血之经，阳气最盛，用刮痧和刺络的方法，最善祛体内热毒。若平日常常敲打，可清洁血液通道，预防青春痘。大肠经对现代医学所讲的淋巴系统有自然保护功能，经常刺激可增强人体免疫力，防止淋巴结核病的生成。) t- F; L) @’ ?+ o5 下面说说这条经络里面的几位“隐士高人”。
三间（俞木穴），位于食指近拇指侧根部，第二掌指关节后。此穴最大的特点就是穴位好找，按摩方便，随时都可以操作。三间穴，最善通经行气，上可通达头面，治疗三叉神经痛、齿痛、目痛、喉肿痛和肩膀痛；下能通腹行气，泻泄可止，便秘可通。另外，有研究指出此穴有消炎、止痛、抗过敏的功效。三间可作为日常的保健穴，常揉多按。本人常用大拇指内侧指节横向硌揉此穴，效果甚佳。 – I; U7 \4 s {& _2 [
阳溪穴（经火穴），翘起拇指，拇指根与背腕之间有一凹陷，凹陷处即为此穴。此穴最善缓解头痛及眼痛酸胀，但若用按摩法，一定要闭目，掐按一分钟，才能有效。此穴名为阳溪，是指阳气像溪水般周流不止，所以此穴最善通经活络，经常按摩，并配合金鸡独立，可以有效防止脑中风和高烧不退等症。 $ v7 w5 l: d” y” ^* w! h( T
手三里，曲肘取穴，在肘横纹头下２寸。提起足三里，向来声名显赫，而手三里却默默无闻，其实经络歌诀中“肚腹三里留”，这个三里，也包括手三里在内的。此穴也善治胃肠病，与足三里并用，效果更佳。此穴还善治腰膝痛，不论是急性慢性，都可点按此穴，可即时缓解症状。手三里善消肿止痛，对于头面肿、牙龈肿、肩臂肿都有疗效。此外手三里还是治疗鼻炎的要穴。手三里可增强体质，是人体的强壮穴，所以平日也可多揉以健身。
大肠经似乎每个穴都有其独特的杀手锏，曲池是治痒奇侠，通治各种皮肤病，还能降血压；偏历善消水肿；肩最散风寒；臂能除眼疾，常点揉此穴可预防白内障，还能治疗麦粒肿。
不被重视的经络也依然是光彩夺目，看来这世间没有什么能隐藏的宝贝，我只是草草地选了几个，匆匆地向大家展示了一下。您若觉得好，就要自己亲手去挑，找您想要的带回去。