absorbing-set · research · also LDPC absorbing set

Absorbing set

An absorbing set is a selected group of variable nodes in a binary parity-check graph where each has more even-degree than odd-degree neighboring checks relative to the group.

Related Observatory article: Read on hesela.com

For a selected nonempty set D, classify each neighboring check by the parity of its number of edges to D. Every selected variable must have strictly more neighbors in the even class than in the odd class. The notation (a,b) counts selected variables and odd neighboring checks, respectively. This is a property of a specified graph, not a universal error-rate formula. [1]

With matrix rows 1100, 1010, 0110 and 1001, select columns 1-3. Their even/odd neighbor counts are 2/1, 2/0 and 2/0, so D is a (3,1) absorbing set. Column 4 has counts 0/1: the set is not fully absorbing. The last row has one selected neighbor, so it is not a stopping set either. This is a synthetic exact example, not a NAND experiment. [2]

In the majority bit-flipping interpretation, treat D as the erroneous positions relative to the all-zero codeword. Selected bits have a local majority of satisfied checks, but outside bits may still flip. A fully absorbing set adds a strict majority condition for every outside bit. The case b=0 is possible; an absorbing set need not be a non-codeword. [1]

The selected absorbing support need not be the eventual failure support of a particular decoder. In the four-column example above, synchronous strict-majority flipping sends 1110 to the wrong codeword 1111. The trapping-set record contrasts that trajectory with likelihood-based decoding. [2]

Relations

related
Fully absorbing set, LDPC (NAND ECC), Stopping set, Trapping set

Citations

  1. Dolecek et al. Analysis of Absorbing Sets and Fully Absorbing Sets of Array-Based LDPC Codes (2010), Definition 1. DOI 10.1109/TIT.2009.2034781. (accessed 2026-11-22)
  2. Hesela: exact synthetic absorbing-set witnesses and executable parity-count model (accessed 2026-11-22)
Sources & caveats
  • Binary graph examples use exact parity arithmetic. They do not reproduce soft-message decoding, channel noise, NAND failure rates or controller behavior.

JSON: /data/terms/absorbing-set.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Absorbing set. Hesela. 2026-11-22. https://hesela.dev/terms/absorbing-set/