Storing Sequences in Binary Tournament-Based Neural Networks

On Neural Associative Memory Structures: Storage and Retrieval of Sequences in a Chain of Tournaments

Associative memories enjoy many interesting properties in terms of error correction capabilities, robustness to noise, storage capacity, and retrieval performance, and their usage spans over a large set of applications. In this letter, we investigate and extend tournament-based neural networks, originally proposed by Jiang, Gripon, Berrou, and Rabbat (2016), a novel sequence storage associative memory architecture with high memory efficiency and accurate sequence retrieval. We propose a more general method for learning the sequences, which we call feedback tournament-based neural networks. The retrieval process is also extended to both directions: forward and backward-in other words, any large-enough segment of a sequence can produce the whole sequence. Furthermore, two retrieval algorithms, cache-winner and explore-winner, are introduced to increase the retrieval performance. Through simulation results, we shed light on the strengths and weaknesses of each algorithm.

Memory integration: An alternative to the consolidation/reconsolidation hypothesis

The original concept of consolidation considers that memory requires time to be fixed. Since 2000, a comparable protein-dependent re-stabilization phase, called reconsolidation, has been assumed to take place after memory retrieval. This consolidation/reconsolidation hypothesis, has dominated the literature for more than 50 years, despite compelling evidence that is inconsistent with it. In this review, we present an historical overview and explain how, despite serious criticisms, this hypothesis has persisted for decades and become accepted as a dogma. Based on both older and more recent evidence, we next propose the concept of memory integration which involves the linkage or embedding of new material into an already existing representation. We believe integration provides a viable explanation for retrograde amnesia in place of the consolidation/reconsolidation hypothesis. Integration can further be the basis for several major cases of memory alteration such as time dependent memory enhancement, interference, counter-conditioning, updating and other instances of memory malleability. In a final section we consider the implications this new concept may have for memory processes and its translational applications.

Nested-Clique Network Model of Neural Associative Memory

Clique-based neural associative memories introduced by Gripon and Berrou (GB), have been shown to have good performance, and in our previous work we improved the learning capacity and retrieval rate by local coding and precoding in the presence of partial erasures. We now take a step forward and consider nested-clique graph structures for the network. The GB model stores patterns as small cliques, and we here replace these by nested cliques. Simulation results show that the nested-clique structure enhances the clique-based model.