A neural network model of when to retrieve and encode episodic memories

  1. Qihong Lu  Is a corresponding author
  2. Uri Hasson
  3. Kenneth A Norman
  1. Princeton University, United States

Abstract

Recent human behavioral and neuroimaging results suggest that people are selective in when they encode and retrieve episodic memories. To explain these findings, we trained a memory-augmented neural network to use its episodic memory to support prediction of upcoming states in an environment where past situations sometimes reoccur. We found that the network learned to retrieve selectively as a function of several factors, including its uncertainty about the upcoming state. Additionally, we found that selectively encoding episodic memories at the end of an event (but not mid-event) led to better subsequent prediction performance. In all of these cases, the benefits of selective retrieval and encoding can be explained in terms of reducing the risk of retrieving irrelevant memories. Overall, these modeling results provide a resource-rational account of why episodic retrieval and encoding should be selective and lead to several testable predictions.

Data availability

The code is made publicly available here in a git repo: https://github.com/qihongl/learn-hippoUsers can also use this code ocean capsule to play with one example model to qualitatively replicate some results: https://codeocean.com/capsule/3639589/tree

The following data sets were generated

Article and author information

Author details

  1. Qihong Lu

    Department of Psychology, Princeton University, Princeton, United States
    For correspondence
    qlu@princeton.edu
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0730-5240
  2. Uri Hasson

    Princeton Neuroscience Institute, Princeton University, Princeton, United States
    Competing interests
    The authors declare that no competing interests exist.
  3. Kenneth A Norman

    Princeton Neuroscience Institute, Princeton University, Princeton, United States
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5887-9682

Funding

Office of Naval Research (Multi-University Research Initiative Grant,ONR/DoD N00014-17-1-2961)

  • Uri Hasson
  • Kenneth A Norman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Copyright

© 2022, Lu et al.

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 5,962
    views
  • 1,090
    downloads
  • 28
    citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Qihong Lu
  2. Uri Hasson
  3. Kenneth A Norman
(2022)
A neural network model of when to retrieve and encode episodic memories
eLife 11:e74445.
https://doi.org/10.7554/eLife.74445

Share this article

https://doi.org/10.7554/eLife.74445

Further reading

    1. Neuroscience
    Lian Hollander-Cohen, Omer Cohen ... Berta Levavi-Sivan
    Research Article

    Life histories of oviparous species dictate high metabolic investment in the process of gonadal development leading to ovulation. In vertebrates, these two distinct processes are controlled by the gonadotropins follicle-stimulating hormone (FSH) and luteinizing hormone (LH), respectively. While it was suggested that a common secretagogue, gonadotropin-releasing hormone (GnRH), oversees both functions, the generation of loss-of-function fish challenged this view. Here, we reveal that the satiety hormone cholecystokinin (CCK) is the primary regulator of this axis in zebrafish. We found that FSH cells express a CCK receptor, and our findings demonstrate that mutating this receptor results in a severe hindrance to ovarian development. Additionally, it causes a complete shutdown of both gonadotropins secretion. Using in-vivo and ex-vivo calcium imaging of gonadotrophs, we show that GnRH predominantly activates LH cells, whereas FSH cells respond to CCK stimulation, designating CCK as the bona fide FSH secretagogue. These findings indicate that the control of gametogenesis in fish was placed under different neural circuits, that are gated by CCK.

    1. Neuroscience
    Lina María Jaime Tobón, Tobias Moser
    Research Article

    Neural diversity can expand the encoding capacity of a circuitry. A striking example of diverse structure and function is presented by the afferent synapses between inner hair cells (IHCs) and spiral ganglion neurons (SGNs) in the cochlea. Presynaptic active zones at the pillar IHC side activate at lower IHC potentials than those of the modiolar side that have more presynaptic Ca2+ channels. The postsynaptic SGNs differ in their spontaneous firing rates, sound thresholds, and operating ranges. While a causal relationship between synaptic heterogeneity and neural response diversity seems likely, experimental evidence linking synaptic and SGN physiology has remained difficult to obtain. Here, we aimed at bridging this gap by ex vivo paired recordings of murine IHCs and postsynaptic SGN boutons with stimuli and conditions aimed to mimic those of in vivo SGN characterization. Synapses with high spontaneous rate of release (SR) were found predominantly on the pillar side of the IHC. These high SR synapses had larger and more temporally compact spontaneous EPSCs, lower voltage thresholds, tighter coupling of Ca2+ channels and vesicular release sites, shorter response latencies, and higher initial release rates. This study indicates that synaptic heterogeneity in IHCs directly contributes to the diversity of spontaneous and sound-evoked firing of SGNs.