文档介绍：Distributed Regression: an Efficient Framework
for Modeling work Data
Carlos Guestrin† Peter Bodik‡ Romain Thibaux‡ Mark Paskin‡ Samuel Madden†
†Intel Research - Berkeley Lab
‡Computer Science Division, University of California, Berkeley
ABSTRACT 1. INTRODUCTION
We present distributed regression, an efficient and works of small, low-power devices that can sense, actu-
framework for work modeling of sensor data. In this ate, municate information about their environment
framework, the nodes of the work collaborate to are proving a useful tool for many tasks in science and indus-
optimally fit a global function to each of their local measure- try. In recent years, developments in hardware and low-level
ments. The algorithm is based upon kernel linear regression, software have led to viable, multi-hundred works
being deployed for remote monitoring of environmental and
where the model takes the form of a weighted sum of lo- +
cal basis functions; this provides an expressive yet tractable climatological data (Mic03; MPS 02).
class of models for work data. Rather than trans- Such monitoring systems are typically used in one of two
modes of operation: either the data from the sensors is ex-
mitting data to one another or outside work, nodes +
communicate constraints on the model parameters, drasti- tracted from work and analyzed off-line (MPS 02);
cally reducing muni