Hippocampal neurons represent events as transferable units of experience

Sun, Chen; Yang, Weijia; Martin, Jared; Tonegawa, Susumu

doi:10.1038/s41593-020-0614-x

Cited by 96 publications

(136 citation statements)

References 47 publications

(71 reference statements)

Supporting

Mentioning

114

Contrasting

Order By: Relevance

“…TEM's medial entorhinal cells have reconfigured to code differently for each lap, understanding that the abstract task space is not a single lap but four ( Figure 5C bottom). This mechanism is consistent with empirical data as manipulation of entorhinal cortex disrupts lap-sensitive hippocampal cells (Sun et al, 2020). However, TEM's medial entorhinal representations stand as a prediction as Sun et al (2020) did not record entorhinal cells.…”

Section: Tem Learned Representations Reflect Transition Statisticssupporting

confidence: 86%

“…To illustrate this, we consider a recent finding by Sun et al (2020). Rodents perform laps of a circular track but only receive reward every four laps.…”

Section: Tem Learned Representations Reflect Transition Statisticsmentioning

confidence: 99%

“…(E) TEM says, however, cells will retain their lap specificity despite spatially remapping (i.e., a lap 2 cell stays a lap 2 cell); since sensory observations repeat each lap, lap specificity is driven by MEC input. (F and G) Analysis to show TEM (F) and real (Sun et al, 2020) (G) hippocampal cells retain lap specificity after remapping. Left: distribution of lap-specificity correlations are significantly higher than shuffles.…”

Section: Structural Knowledge Is Preserved Over Apparently Random Hipmentioning

confidence: 99%

See 2 more Smart Citations

The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation

Whittington

Muller

Mark

et al. 2020

Cell

365

396

View full text Add to dashboard Cite

Summary The hippocampal-entorhinal system is important for spatial and relational memory tasks. We formally link these domains, provide a mechanistic understanding of the hippocampal role in generalization, and offer unifying principles underlying many entorhinal and hippocampal cell types. We propose medial entorhinal cells form a basis describing structural knowledge, and hippocampal cells link this basis with sensory representations. Adopting these principles, we introduce the Tolman-Eichenbaum machine (TEM). After learning, TEM entorhinal cells display diverse properties resembling apparently bespoke spatial responses, such as grid, band, border, and object-vector cells. TEM hippocampal cells include place and landmark cells that remap between environments. Crucially, TEM also aligns with empirically recorded representations in complex non-spatial tasks. TEM also generates predictions that hippocampal remapping is not random as previously believed; rather, structural knowledge is preserved across environments. We confirm this structural transfer over remapping in simultaneously recorded place and grid cells.

show abstract

Section: Tem Learned Representations Reflect Transition Statisticssupporting

confidence: 86%

“…To illustrate this, we consider a recent finding by Sun et al (2020). Rodents perform laps of a circular track but only receive reward every four laps.…”

Section: Tem Learned Representations Reflect Transition Statisticsmentioning

confidence: 99%

Section: Structural Knowledge Is Preserved Over Apparently Random Hipmentioning

confidence: 99%

See 1 more Smart Citation

The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation

Whittington

Muller

Mark

et al. 2020

Cell

365

396

View full text Add to dashboard Cite

show abstract

“…For example, one representation of interest might be whatever is common at a particular location in space but invariant to what sensory stimuli are present at that location 15 . A related type of abstraction corresponds to the position of an item in a sequence, invariant to which actual item is in that position 12,16 .…”

Section: States As Abstractionsmentioning

confidence: 99%

Measuring Sequences of Representations with Temporally Delayed Linear Modelling

Liu

Dolan

Penagos-Vargas

et al. 2020

Preprint

View full text Add to dashboard Cite

There are rich structures in off-task neural activity. For example, task related neural codes are thought to be reactivated in a systematic way during rest. This reactivation is hypothesised to reflect a fundamental computation that supports a variety of cognitive functions. Here, we introduce an analysis toolkit (TDLM) for analysing this activity. TDLM combines nonlinear classification and linear temporal modelling to testing for statistical regularities in sequences of neural representations. It is developed using non-invasive neuroimaging data and is designed to take care of confounds and maximize sequence detection ability. The method can be extended to rodent electrophysiological recordings. We outline how TDLM can successfully reveal human replay during rest, based upon non-invasive magnetoencephalography (MEG) measurements, with strong parallels to rodent hippocampal replay. TDLM can therefore advance our understanding of sequential computation and promote a richer convergence between animal and human neuroscience research.TDLM is a general, and flexible, tool for measuring neural sequences. It facilitates crossspecies investigations by linking large-scale measurements in humans to cellular measurements in non-human species. We outline its promise for revealing abstract cognitive processes that extend beyond sensory representation, potentially opened doors for new avenues of research in cognitive science. All code and facilities will be available at https://github.com/yunzheliu/TDLM. RESULTS TDLM Overview of TDLMOur primary goal is to test for temporal structure in neural activity. To achieve this, we would like ideally a method which (1) uncovers regularity in the reactivation of neural activity, (2) tests whether this regularity conforms to a hypothesized structure. Here the structure between neural representation is expressed as their sequential reactivation in time, i.e., sequence. In what follows, we will use the terms "temporal structure" and "sequence" interchangeably.The starting point of TDLM is a set of n time series, each corresponding to a decoded neural representation of a variable of interest. These time series could themselves be obtained in several ways, described in detail in a later section ("Getting the states"). The aim of TDLM is to identify task-related regularities in sequences of these representations off-task.Consider, for example, a task in which participants have been trained such that n=4 distinct sensory cues (A, B, C, and D) appear in a consistent order ( → → → ) (Fig 1a). If we are interested in replay of this sequence during subsequent resting periods, we might want to ask statistical questions of the following form: "Does the existence of a neural representation of A, at time T in the rest period, predict the occurrence of a representation of B at time T+∆ ", and similarly for → and → .In TDLM we ask such questions using a two-step process. First, for each of the n 2 possible pairs of variables Xi and Xj, we find the correlation between the Xi time series and the ∆ -shifted...

show abstract

“…continuous random variables. The KSG estimator has been used to estimate behavioral information 55 contained in continuous neural data, such as FMRI (Gómez-Verdejo et al, 2012), AP waveform shape 56 , MEG (Robinson and Mandell, 2015), LFP (Manyakov and Van Hulle, 2008), and 57 intracellular recordings (Sun et al, 2020;Wang et al, 2017). 58…”

Section: Introduction 18mentioning

confidence: 99%

Information theoretic approaches to deciphering the neural code with functional fluorescence imaging

Climer

Dombeck

2020

Preprint

View full text Add to dashboard Cite

6 Information theoretic metrics have proven highly useful to quantify the relationship between behaviorally 7 relevant parameters and neuronal activity with relatively few assumptions. However, such metrics are 8 typically applied to action potential recordings and were not designed for the slow timescales and variable 9 amplitudes typical of functional fluorescence recordings (e.g. calcium imaging). Therefore, the power of 10 information theoretic metrics has yet to be fully exploited by the neuroscience community due to a lack 11 of understanding for how to apply and interpret the metrics with such fluorescence traces. Here, we used 12 computational methods to create mock action potential traces with known amounts of information and 13 from them generated fluorescence traces to examine the ability of different information metrics to recover 14 the known information values. We provide guidelines for the use of information metrics when applied to 15 functional fluorescence and demonstrate their appropriate application to GCaMP6f population recordings 16 from hippocampal neurons imaged during virtual navigation. 17

show abstract

Hippocampal neurons represent events as transferable units of experience

Cited by 96 publications

References 47 publications

The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation

The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation

Measuring Sequences of Representations with Temporally Delayed Linear Modelling

Information theoretic approaches to deciphering the neural code with functional fluorescence imaging

Contact Info

Product

Resources

About