Flexible Kernel Memory
Dimitri Nowicki, Hava Siegelmann · 2010
Read Flexible Kernel Memory by Dimitri Nowicki, Hava Siegelmann (2010) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0010955.txt.
Description
Dimitri Nowicki and Hava Siegelmann published this computing study in PLOS ONE in 2010. Associative memory networks store patterns and retrieve them from partial cues, but classic models struggle when inputs change size or when memories must be added without retraining. The authors introduce a flexible kernel memory that handles both yes-no and continuous inputs.
Built on kernel theory, the model behaves like a neural memory in a transformed feature space, and its stored attractors can be added, deleted, or updated online without harming older ones. Inputs may grow or shrink in dimension without relearning, and a consolidation process groups inputs into clusters, demonstrated on handwritten digits. Nowicki and Siegelmann even add a reconsolidation-like step echoing human memory. Their design points toward memory systems that keep learning without forgetting.
Catalog record
| CREATOR | Dimitri Nowicki, Hava Siegelmann |
|---|---|
| YEAR | 2010 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| LICENSE | Creative Commons Attribution License |
| RIGHTS | CC BY |
| SOURCE | PLOS ONE |
Source: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0010955