An attractor network model of serial recall

2010

Traditionally, cognitive neuroscientists have represented long-term memory in terms of the structure of a neural network's connections and short-term memory in terms of the patterns of activation across the network (e.g., . However, recent neural-network models of short-term verbal working memory (VWM) have used modifiable structural connections to encode item and order information . In these models, words are stored by changing the connection weights between linguistic units, and phenomena related to VWM are thus modeled with long-term memory structures and mechanisms. Although it may be possible for neural connections to change rapidly (e.g., see , this latter approach to modeling VWM does not appear to be motivated by neurobiology. Consequently, we have formulated a neural-network model of VWM that stores and maintains item and order information as a pattern of activation within a fixed-structure network whose connections remain constant. Performance with this network is affected by many of the standard factors that influence VWM performance, such as phonological similarity, articulatory duration, and word frequency. Our work demonstrates that the embodiment of short-term memory based on patterns of network activation is feasible, parsimonious, and merits more investigation. Furthermore, from our work it appears that both the biology and psychology of specific mental processes must be understood more fully before neural networks can be deemed "neurally plausible".

PLoS ONE, 2013

The dynamic nature of human working memory, the general-purpose system for processing continuous input, while keeping no longer externally available information active in the background, is well captured in immediate free recall of supraspan word-lists. Free recall tasks produce several benchmark memory phenomena, like the U-shaped serial position curve, reflecting enhanced memory for early and late list items. To account for empirical data, including primacy and recency as well as contiguity effects, we propose here a neurobiologically based neural network model that unifies shortand long-term forms of memory and challenges both the standard view of working memory as persistent activity and dualstore accounts of free recall. Rapidly expressed and volatile synaptic plasticity, modulated intrinsic excitability, and spikefrequency adaptation are suggested as key cellular mechanisms underlying working memory encoding, reactivation and recall. Recent findings on the synaptic and molecular mechanisms behind early LTP and on spiking activity during delayedmatch-to-sample tasks support this view.

2002

Abstract The class of recurrent networks known as attractor networks is known to exhibit behaviors relevant to modeling human memory processes���notably content-addressable memory, storage of repeated inputs as stable patterns (under Hebbian learning), and maintenance of information (as activity) over time. In addition, these networks provide a natural account of the effect of similarity on interference in recall.

1993

We study an Attractor Neural Network that stores natural concepts, organized in semantic classes. The concepts are represented by distributed patterns over a space of attributes, and are related by both semantic and episodic associations. While semantic relations are expressed through an hierarchical coding over the attribute space, episodic links are realized via specific synaptic projections.

Frontiers in Computational Neuroscience, 2015

Human memory can store large amount of information. Nevertheless, recalling is often a challenging task. In a classical free recall paradigm, where participants are asked to repeat a briefly presented list of words, people make mistakes for lists as short as 5 words. We present a model for memory retrieval based on a Hopfield neural network where transition between items are determined by similarities in their long-term memory representations. Meanfield analysis of the model reveals stable states of the network corresponding (1) to single memory representations and (2) intersection between memory representations. We show that oscillating feedback inhibition in the presence of noise induces transitions between these states triggering the retrieval of different memories. The network dynamics qualitatively predicts the distribution of time intervals required to recall new memory items observed in experiments. It shows that items having larger number of neurons in their representation are statistically easier to recall and reveals possible bottlenecks in our ability of retrieving memories. Overall, we propose a neural network model of information retrieval broadly compatible with experimental observations and is consistent with our recent graphical model (Romani et al., 2013).

Journal of Memory and Language, 2006

The interaction between short-and long-term memory is studied within a model in which phonemic and (temporal) contextual information have separate influences on immediate verbal serial recall via connections with short-and longterm plasticity [Burgess, N., & Hitch, G.J. (1999). Memory for serial order: a network model of the phonological loop and its timing. Psychological Review, 106, 551-581]. Long-term learning of sequences of familiar items is correctly predicted to interact with temporal grouping but not phonological similarity or articulatory suppression. However the model fails to predict learning of different sequences simultaneously, or of partially repeated lists. In a revised model, sufficiently different sequences recruit different context signals while sufficiently similar sequences recruit the same signal, via a cumulative matching process during encoding. Simulations show this revised model captures the experimental data on Hebb repetition, including the importance of matching at the start of a list, makes novel predictions concerning the effects of partial repetition, and provides a potential mechanism for position specific intrusions and the build up of proactive interference.

2017

Working memory, the ability to keep recently encountered information available for immediate processing, has been proposed to rely on two mechanisms that appear difficult to reconcile: self-sustained neural firing, or the opposite — activity-silent synaptic traces. Here we show that both phenomena can co-exist within a unified system in which neurons hold information in both activity and synapses. Rapid plasticity in flexibly-coding neurons allows features to be bound together into objects, with an important emergent property being the focus of attention. One memory item is held by persistent activity in an attended or “focused” state, and is thus remembered better than other items. Other, previously attended items can remain in memory but in the background, encoded in activity-silent synaptic traces. This dual functional architecture provides a unified common mechanism accounting for a diverse range of perplexing attention and memory effects that have been hitherto difficult to exp...

Studies on the task of serial recall in the verbal domain have discovered several phenomena, raising questions about how memory encodes and recalls data. Quantitative models of verbal short term memory try to answer these questions by providing computational representations of certain memory aspects. This paper introduces different serial recall models by addressing their main principles and limitations. It is concluded that currently none of these models is able to explain all phenomena associated with serial recall and that the problem of verbal short term memory might be solved best via a unifying approach that consolidates core assumptions of different existing models.

[Proceedings 1992] IJCNN International Joint Conference on Neural Networks, 1992

A working memory model is describ<~d that is capable of storing and recalling arbitrary temporal sequences of events, including repeated items. These memories encode the invariant temporal order of sequential events that may be present<~d at widely differing speeds, durations, and interstimulus intervals. This temporal order code is designed to enable all possible groupings of sequential events to be stably learned and re1nembered in real ti1ne, even as new events perturb the syste1n.