How Does the Brain Solve the Computational Problems of Spatial Navigation?

Widloski, John; Fiete, Ila

doi:10.1007/978-3-7091-1292-2_14

Cited by 8 publications

(8 citation statements)

References 221 publications

(303 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Whereas animal studies and theoretical models have long suggested that path integration is a key function of the grid cell system, direct empirical evidence for this claim is scarce [ 4 , 5 , 6 , 25 , 31 , 32 , 33 , 34 , 35 , 36 ]. Our data show that the magnitude of grid-cell-like representations in the human entorhinal cortex is linked to path integration performance in old age and therefore further strengthens the hypothesis that grid cell function underlies path integration processes.…”

Section: Resultsmentioning

confidence: 99%

“… Summary A progressive loss of navigational abilities in old age has been observed in numerous studies, but we have only limited understanding of the neural mechanisms underlying this decline [ 1 ]. A central component of the brain’s navigation circuit are grid cells in entorhinal cortex [ 2 ], largely thought to support intrinsic self-motion-related computations, such as path integration (i.e., keeping track of one’s position by integrating self-motion cues) [ 3 , 4 , 5 , 6 ]. Given that entorhinal cortex is particularly vulnerable to neurodegenerative processes during aging and Alzheimer’s disease [ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 ], deficits in grid cell function could be a key mechanism to explain age-related navigational decline.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Compromised Grid-Cell-like Representations in Old Age as a Key Mechanism to Explain Age-Related Navigational Deficits

et al. 2018

View full text Add to dashboard Cite

SummaryA progressive loss of navigational abilities in old age has been observed in numerous studies, but we have only limited understanding of the neural mechanisms underlying this decline [1]. A central component of the brain’s navigation circuit are grid cells in entorhinal cortex [2], largely thought to support intrinsic self-motion-related computations, such as path integration (i.e., keeping track of one’s position by integrating self-motion cues) [3, 4, 5, 6]. Given that entorhinal cortex is particularly vulnerable to neurodegenerative processes during aging and Alzheimer’s disease [7, 8, 9, 10, 11, 12, 13, 14], deficits in grid cell function could be a key mechanism to explain age-related navigational decline. To test this hypothesis, we conducted two experiments in healthy young and older adults. First, in an fMRI experiment, we found significantly reduced grid-cell-like representations in entorhinal cortex of older adults. Second, in a behavioral path integration experiment, older adults showed deficits in computations of self-position during path integration based on body-based or visual self-motion cues. Most strikingly, we found that these path integration deficits in older adults could be explained by their individual magnitudes of grid-cell-like representations, as reduced grid-cell-like representations were associated with larger path integration errors. Together, these results show that grid-cell-like representations in entorhinal cortex are compromised in healthy aging. Furthermore, the association between grid-cell-like representations and path integration performance in old age supports the notion that grid cells underlie path integration processes. We therefore conclude that impaired grid cell function may play a key role in age-related decline of specific higher-order cognitive functions, such as spatial navigation.

show abstract

Section: Resultsmentioning

confidence: 99%

mentioning

confidence: 99%

Compromised Grid-Cell-like Representations in Old Age as a Key Mechanism to Explain Age-Related Navigational Deficits

et al. 2018

View full text Add to dashboard Cite

show abstract

“…2b -Relationship to previous models of grid cells Our work contrasts with previous approaches where grid cells have been hard-wired 53-56 and 57 , derived through eigendecomposition of place fields 58,59 , or arisen through self organization in the absence of an objective function 60 . It is worth noting that our experiments were not designed to provide insights into the development of grid cells in the brain -due to the limitations of the training algorithm used (i.e.…”

Section: A3cmentioning

confidence: 94%

Vector-based navigation using grid-like representations in artificial agents

Banino

Barry

Uría

et al. 2018

Nature

515

579

View full text Add to dashboard Cite

Deep neural networks have achieved impressive successes in fields ranging from object recognition to complex games such as Go. Navigation, however, remains a substantial challenge for artificial agents, with deep neural networks trained by reinforcement learning failing to rival the proficiency of mammalian spatial behaviour, which is underpinned by grid cells in the entorhinal cortex . Grid cells are thought to provide a multi-scale periodic representation that functions as a metric for coding space and is critical for integrating self-motion (path integration) and planning direct trajectories to goals (vector-based navigation). Here we set out to leverage the computational functions of grid cells to develop a deep reinforcement learning agent with mammal-like navigational abilities. We first trained a recurrent network to perform path integration, leading to the emergence of representations resembling grid cells, as well as other entorhinal cell types . We then showed that this representation provided an effective basis for an agent to locate goals in challenging, unfamiliar, and changeable environments-optimizing the primary objective of navigation through deep reinforcement learning. The performance of agents endowed with grid-like representations surpassed that of an expert human and comparison agents, with the metric quantities necessary for vector-based navigation derived from grid-like units within the network. Furthermore, grid-like representations enabled agents to conduct shortcut behaviours reminiscent of those performed by mammals. Our findings show that emergent grid-like representations furnish agents with a Euclidean spatial metric and associated vector operations, providing a foundation for proficient navigation. As such, our results support neuroscientific theories that see grid cells as critical for vector-based navigation, demonstrating that the latter can be combined with path-based strategies to support navigation in challenging environments.

show abstract

“…However, when external sensory signals are not present or are spatially uninformative, place cells nevertheless respond reliably and repeatably O' Keefe (1976); Quirk et al (1990). Interpreting in a neural setting the classic ideas of cognitive maps O' Keefe and Nadel (1978); Tolman (1948); McNaughton et al (2006) and models of simultaneous localization and mapping (SLAM) Leonard and Durrant-Whyte (1991); Milford et al (2004); Cadena et al (2016); Cheung et al (2012); Widloski and Fiete (2014); Kanitscheider and Fiete (2017c,b,a), we espouse the view that a likely role of place cells is to develop appropriate bindings (maps) between an internal self-consistent scaffold of spatial coordinate representations (presumably generated by grid cells) and external sensory cues (including rewards in the world). In this view, low-dimensional structured grid responses induce complex place cell activity patterns, rather than the alternate view in which simple place field patterns induce grid responses Dordek et al (2016).…”

Section: Introductionmentioning

confidence: 99%

Where can a place cell put its fields? Let us count the ways

Yim

Sadun

Fiete

et al. 2019

Preprint

View full text Add to dashboard Cite

A hippocampal place cell exhibits multiple firing fields within and across environments. What factors determine the configuration of these fields, and could they be set down in arbitrary locations? We conceptualize place cells as performing evidence combination across many inputs and selecting a threshold to fire. Thus, mathematically they are perceptrons, except that they act on geometrically organized inputs in the form of multiscale periodic grid-cell drive, and external cues. We analytically count which field arrangements a place cell can realize with structured grid inputs, to show that many more place-field arrangements are realizable with grid-like than one-hot coded inputs. However, the arrangements have a rigid structure, defining an underlying response scaffold. We show that the "separating capacity" or spatial range over which all potential field arrangements are realizable equals the rank of the grid-like input matrix, which in turn equals the sum of distinct grid periods, a small fraction of the unique grid-cell coding range. Learning different grid-to-place weights beyond this small range will alter previous arrangements, which could explain the volatility of the place code. However, compared to random inputs over the same range, grid-structured inputs generate larger margins, conferring relative robustness to place fields when grid input weights are fixed. Significance statementPlace cells encode cognitive maps of the world by combining external cues with an internal coordinate scaffold, but our ability to predict basic properties of the code, including where a place cell will exhibit fields without external cues (the scaffold), remains weak. Here we geometrically characterize the place cell scaffold, assuming it is derived from multiperiodic modular grid cell inputs, and provide exact combinatorial results on the space of permitted field arrangements. We show that the modular inputs permit a large number of place field arrangements, with robust fields, but also strongly constrain their geometry and thus predict a structured place scaffold.

show abstract

How Does the Brain Solve the Computational Problems of Spatial Navigation?

Cited by 8 publications

References 221 publications

Compromised Grid-Cell-like Representations in Old Age as a Key Mechanism to Explain Age-Related Navigational Deficits

Compromised Grid-Cell-like Representations in Old Age as a Key Mechanism to Explain Age-Related Navigational Deficits

Vector-based navigation using grid-like representations in artificial agents

Where can a place cell put its fields? Let us count the ways

Contact Info

Product

Resources

About