Learning the Vector Coding of Egocentric Boundary Cells from Visual Data

Lian, Yanbo; Williams, Simon; Alexander, Andrew S.; Hasselmo, Michael E.; Burkitt, Anthony N.

doi:10.1101/2022.01.28.478267

Cited by 3 publications

(3 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, the fact that these egocentric boundary cells can respond at distances up to 50 cm from the boundary wall suggest that they are guided by visual sensory input (Alexander et al, 2020). Computational models show that using synaptic modification with sparse coding constraints (Lian & Burkitt, 2021; Olshausen & Field, 1996, 2004) to form an efficient representation of visual input from the entire foraging trajectory within a population results in many simulated neurons that have responses similar to egocentric boundary cells that can respond similarly in circular and rectangular environments (Lian et al, 2022).…”

Section: Review Of Models Of Spatial Coordinate Representationsmentioning

confidence: 99%

Gated transformations from egocentric to allocentric reference frames involving retrosplenial cortex, entorhinal cortex, and hippocampus

et al. 2023

View full text Add to dashboard Cite

This paper reviews the recent experimental finding that neurons in behaving rodents show egocentric coding of the environment in a number of structures associated with the hippocampus. Many animals generating behavior on the basis of sensory input must deal with the transformation of coordinates from the egocentric position of sensory input relative to the animal, into an allocentric framework concerning the position of multiple goals and objects relative to each other in the environment. Neurons in retrosplenial cortex show egocentric coding of the position of boundaries in relation to an animal. These neuronal responses are discussed in relation to existing models of the transformation from egocentric to allocentric coordinates using gain fields and a new model proposing transformations of phase coding that differ from current models. The same type of transformations could allow hierarchical representations of complex scenes. The responses in rodents are also discussed in comparison to work on coordinate transformations in humans and non-human primates.

show abstract

Section: Review Of Models Of Spatial Coordinate Representationsmentioning

confidence: 99%

Gated transformations from egocentric to allocentric reference frames involving retrosplenial cortex, entorhinal cortex, and hippocampus

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The sparse coding model used in this paper is implemented using locally competitive algorithm (Rozell et al, 2008), which is also used in other works (Zhu and Rozell, 2013; Lian and Burkitt, 2021, 2022; Lian et al, 2023). Though some components of the model are not biologically plausible, the essence of this paper is to demonstrate that the principle of sparse coding can be used to account for the development of spatiotemporal properties V1 cells.…”

Section: Discussionmentioning

confidence: 99%

Learning spatio-temporal V1 cells from diverse LGN inputs

Ruslim,

Burkitt,

Lian

2023

Preprint

Self Cite

View full text Add to dashboard Cite

Since the Nobel-winning discovery of simple cells and complex cells in cat’s primary visual cortex (V1) by Hubel and Wiesel, many experimental studies of the visual system of the brain have been conducted. Experimental data of V1 cells from animal recordings show spatio-temporal properties, namely that they display both spatial and temporal response properties. For spatial properties, each V1 cell responds to a specific feature (such as bars, blobs, etc.) in the visual space, which is called the receptive field of this cell. The receptive fields of different V1 cells are typically different in size, orientation, spatial frequencies, etc. Furthermore, V1 cells also display temporal properties, namely that the receptive fields of V1 cells can change over time, and there is a great variety of ways in which they change over time. For example, the population of V1 cells show a great diversity from monophasic response to biphasic response, and some V1 cells are selective to a preferred direction. However, given many computational learning models that explain how spatial properties of V1 cells can be learnt, how temporal properties emerge is still not well understood. In this paper, we use a simple learning model based on sparse coding to show that spatio-temporal V1 cells, such as biphasic and direction selective cell, can emerge via synaptic plasticity when diverse spatio-temporal LGN cells are used as upstream input to V1 cells. This work suggests that temporal along with spatial properties of V1 cells may simply come from a learning process that aims to encode upstream input with spatio-temporal properties, which greatly enhances our understanding of V1 cells.

show abstract

“…An appealing feature of sparse coding is its learning ability, namely that the model describes how the connection weights can be learnt based on the statistics of the input visual data, in a way that is similar to synaptic plasticity. Apart from its success in learning RFs of cells in the visual cortex, sparse coding has also been successfully applied to explain experimental results in other brain areas, such as auditory cortex (Smith and Lewicki, 2006) and hippocampal formation (Lian and Burkitt, 2021, 2022; Lian et al, 2023). For a review see Beyeler et al (2019).…”

Section: Introductionmentioning

confidence: 99%

Unifying sparse coding, predictive coding, and divisive normalization

Lian

Burkitt

2023

Preprint

Self Cite

View full text Add to dashboard Cite

Sparse coding, predictive coding and divisive normalization are thought to be underlying principles of neural circuits across the brain, supported by much experimental evidence. However, the connection among these principles is still poorly understood. In this paper, we aim to unify sparse coding, predictive coding, and divisive normalization in the context of a two-layer neural model. This two-layer model is constructed that implements sparse coding with the structure originating from predictive coding. Our results show that a homeostatic function in the model can shape the nonlinearity of neural responses, which can replicate different forms of divisive normalization. We demonstrate that this model is equivalent to divisive normalization in a single-neuron scenario. Our simulation also shows that the model can learn simple cells with the property of contrast saturation that is previously explained by divisive normalization. In summary, our study demonstrates that the three principles of sparse coding, predictive coding, and divisive normalization can be unified, such that the model has the ability to learn as well as display more diverse response nonlinearities. This framework may also be potentially used to explain how the brain learns to integrate input from different sensory modalities.

show abstract

Learning the Vector Coding of Egocentric Boundary Cells from Visual Data

Cited by 3 publications

References 54 publications

Gated transformations from egocentric to allocentric reference frames involving retrosplenial cortex, entorhinal cortex, and hippocampus

Gated transformations from egocentric to allocentric reference frames involving retrosplenial cortex, entorhinal cortex, and hippocampus

Learning spatio-temporal V1 cells from diverse LGN inputs

Unifying sparse coding, predictive coding, and divisive normalization

Contact Info

Product

Resources

About