nucleation

LIT-1848: Sparse Distributed Memory: Principles and Operation

2 min read

Properties4
StatusActive — read 2026-10-30 (NOTE-1454), Skimmed: chapter 1 in full (pp. 1–28) and chapters 2–3 looked over, from NASA's public-domain NTRS copy (ntrs.nasa.gov/api/citations/19920001076/downloads/19920001076.pdf, 60 scanned pages with an OCR text layer). Worth reading as a short, first-hand statement by Kanerva and two co-authors of what SDM's critical distance is: a cue closer than it is pulled toward the stored pattern by iterated reads; one farther away "will not converge" and wanders, possibly into another pattern's attracting zone; the distance shrinks as the memory fills; and in an autoassociative memory the two outcomes are "easy to distinguish in practice". Read it knowing that the critical distance is asserted, not derived (the derivation is referred to Kanerva's book, LIT-928, still unread), that no simulation data are given for the convergence figures, and that most of the report describes the Stanford hardware prototype.
Tagscognition, information-retrieval
Cited as “grounds” byCLAIM-180 — Drift is retrieval failure outside a critical distance: a restatement close enough to an idea retrieves it, one that has drifted too far is taken for new, and a code bypasses retrieval by content altogether
Cited as “paper” byNOTE-1454 — Flynn, Kanerva & Bhadkamkar — SDM principles and operation

Graph View

Flynn, M. J., Kanerva, P. & Bhadkamkar, N. (1989). Sparse Distributed Memory: Principles and Operation. RIACS Technical Report 89.53, NASA Ames Research Center; Stanford CSL-TR-89-400. NTRS 19920001076.

Key takeaways

From the reading, NOTE-1454 (Skimmed: chapter 1 in full).

  • The memory. Words are both addresses and data. A write adds the word, bit by bit as +1 or −1, into the counters of every hard location within Hamming radius d of its address; a read sums the counters of the locations within radius d′ of the cue and thresholds each sum (§1.4.4, pp. 11–14).
  • The critical distance. “If Z is closer than a certain critical distance to P(1), then summing and thresholding … results in a vector Z(1) that is closer to P(1) than Z was”, so iterated reads converge on P(1). “The greater the population of the memory, the smaller the critical distance.” A cue beyond it gives “a diverging sequence that will wander through address space” and may fall into another pattern’s “attracting zone” (§1.5, pp. 15–16).
  • Divergence is visible for patterns, not for sequences. Convergence takes about ten iterations and divergence many, so “the two situations are easy to distinguish in practice” (p. 16); for stored sequences “the difference is unfortunately hard to tell” (p. 18).
  • Not derived here. The convergence claim is “shown mathematically” in Kanerva’s book; the report gives neither the derivation nor data.

Standing in the record

Filed on 2026-10-30 for CLAIM-180, whose statement of the critical distance rests on NOTE-242’s reading of LIT-269 while the book it comes from, LIT-928, is unread. This report is Kanerva’s own short statement of the property, and confirms the claim’s description of it, but it defers the mathematics to the book. Most of the report describes the hardware.

Backlinks