IDEALEM
Implementation of Dynamic Extensible Adaptive Locally Exchangeable Measures

By checking this box, you acknowledge that the software and material on this web site is a non-commercial, academic, non-profit use only. "IDEALEM: Implementation of Dynamic Extensible Adaptive Locally Exchangeable Measures (IDEALEM)" Copyright (c) 2016, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved. BY DOWNLOADING, INSTALLING, OR USING THE SOFTWARE, YOU ARE INDICATING YOUR ACCEPTANCE OF THE TERMS AND CONDITIONS HEREIN. Full text of LICENSE Full text of COPYRIGHT

Before going to download page... The lab needs to collect the software distribution info. Please fill in the information below and press Go! button. Name Institution Email

Summary of IDEALEM (Implementation of Dynamic Extensible Adaptive Locally Exchangeable Measures) Data is dynamically reduced with a novel pattern searching method based on statistical similarity. Compressed/reduced data has accurate representation for the original data, i.e., decompressed/reconstructed data has the same statistical data distribution as the original data. The method supports event/feature detection directly on the compressed data.
Fig. 1	Fig. 2
Fig. 3
Fig. 1 shows micro PMU (Phase Measurement Unit) data from power grid electricity measurement from one of on-site switches at LBNL (in collaboration with Energy Technology Area). In Fig 2, the data compression ratio (original size in bytes / compressed size) in this use case is 95.23 with only one history buffer, and can be achieved using only 64K bytes of memory. Compared to gzip, IDEALEM compressed data size is under 2% of gzip-compressed data size in bytes. Fig. 3 shows another micro PMU data from power grid electricity measurement from one of on-site switches at LBNL (in collaboration with Energy Technology Area), which results the compression ratio of 242.3 with 255 history buffers. IDEALEM is an implementation of the data reduction and pattern searching algorithm for streaming data based on Locally Exchangeable Measures, U.S. Patent 10,366,078. Software is available for commercial and non-commercial use.

Related links:

• Efficient Data Reduction Method with Locally Exchangeable Measures IB-2013-133
• IDEALEM presentation sides (Nov. 14, 2016)
• Supercomputing 2016 demo

Questions or comments:

Credits:

• Relational Dynamic Bayesian Networks with Locally Exchangeable Measures, LBNL Tech Report 6341E
• Novel Data Reduction Based on Statistical Similarity, SSDBM 2016

We gratefully acknowledge the collaboration and helpful suggestions from colleagues and friends for this project. In particular, we thank Emma Stewart, Sean Peisert, Chuck McParland, Reinhard Gentz, Mahdi Jamei, Ciaran Roberts, and Sebastian Ainslie .

Disclaimer
LICENSE
This web page is maintained by the Scientific Data Management Group
Lawrence Berkeley National Laboratory

Last updated: Thu Mar 3 14:11:32 PST 2016