Description:

Letter Salad or Salad for short, is an efficient and flexible implementation of the well-known anomaly detection method Anagram by Wang et al. (RAID 2006) and provides various extensions to it.

Salad is based on n-gram models, that is, data is represented as all its substrings of length n. During training these n-grams are stored in a Bloom filter. This enables the detector to represent a large number of n-grams in little memory and still being able to efficiently access the data. Salad extends Anagram by allowing various n-gram types, a 2-class version of the detector for classification and various model analysis modes.

Changes to previous version:

After a full year of development we proudly present you several new features, plenty of bug fixes and better performance :)

• It now is possible to process data on bit granularity salad [train|inspect] --binary
• Performance improvements while simultaneously preserving and further advancing readability of the source code.
• Suppress the verbose output of Salad salad [train|predict] -q
• Extend the (unit) testing framework to support test of the overall application and memchecks using valgrind.
• Testing mode was renamed: salad dbg -> salad test
• Allow to select either client or server-side data when processing network communication.
• libfoodstoragebox A library encapsulating advanced data structures such as bloom filters.
• Fixes for a critical bug when using group input and several minor issues.
• An optionally compressed, text-based model file format salad train -F (txt|archive)
• The default hashset ('simple2') makes use of djb2 hash
• Flawless builds using gcc, mingw and clang

BibTeX Entry: Download

Corresponding Paper BibTeX Entry: Download

Supported Operating Systems: Linux, Windows, Unix, Posix, Mac Os X

Data Formats: Binary, Txt

Tags: Sequence Analysis, Sparse Learning

Archive: download here

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.