Coordinated drift of receptive fields in Hebbian/anti-Hebbian network models during noisy representation learning

Qin, Shanshan; Farashahi, Shiva; Lipshutz, David; Sengupta, Anirvan M.; Chklovskii, Dmitri B.; Pehlevan, Cengiz

doi:10.1038/s41593-022-01225-z

Cited by 22 publications

(42 citation statements)

References 75 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Long-term recordings of large populations of neurons have revealed changing neural representations over time, a phenomenon termed representational drift 51,52 . This has been documented in the hippocampus 53 as well as several other brain areas [54][55][56][57][58][59] .…”

Section: Discussionmentioning

confidence: 99%

Sub-cellular population imaging tools reveal stable apical dendrites in hippocampal area CA3

Moore

Rashid

Johnson

et al. 2023

Preprint

View full text Add to dashboard Cite

Anatomically segregated apical and basal dendrites of pyramidal neurons receive functionally distinct inputs, but it is unknown if this results in compartment-level functional diversity during behavior. Here we imaged calcium signals from apical dendrites, soma, and basal dendrites of pyramidal neurons in area CA3 of mouse hippocampus during head-fixed navigation. To examine dendritic population activity, we developed computational tools to identify dendritic regions of interest and extract accurate fluorescence traces. We identified robust spatial tuning in apical and basal dendrites, similar to soma, though basal dendrites had reduced activity rates and place field widths. Across days, apical dendrites were more stable than soma or basal dendrites, resulting in better decoding of the animal’s position. These population-level dendritic differences may reflect functionally distinct input streams leading to different dendritic computations in CA3. These tools will facilitate future studies of signal transformations between cellular compartments and their relation to behavior.

show abstract

Section: Discussionmentioning

confidence: 99%

Sub-cellular population imaging tools reveal stable apical dendrites in hippocampal area CA3

Moore

Rashid

Johnson

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The tensor rank of learning in RNNs with local synaptic rules (e.g., Hebbian) remains an open question. Towards this end, theoretical work has established links between Hebbian and gradient-based learning [43,44], opening the possibility of an extension of our mathematical results towards more biologically plausible learning rules.…”

Section: Discussionmentioning

confidence: 99%

Disentangling Mixed Classes of Covariability in Large-Scale Neural Data

Pellegrino

Stein

Cayco-Gajic

2023

Preprint

View full text Add to dashboard Cite

Recent work has argued that large-scale neural recordings are often well described by low-dimensional ‘latent’ dynamics identified using dimensionality reduction. However, the view that task-relevant variability is shared across neurons misses other types of structure underlying behavior, including stereotyped neural sequences or slowly evolving latent spaces. To address this, we introduce a new framework that simultaneously accounts for variability that is shared across neurons, trials, or time. To identify and demix these covariability classes, we develop a new unsupervised dimensionality reduction method for neural data tensors called sliceTCA. In three example datasets, including motor cortical dynamics during a classic reaching task and recent multi-region recordings from the International Brain Laboratory, we show that sliceTCA can capture more task-relevant structure in neural data using fewer components than traditional methods. Overall, our theoretical framework extends the classic view of low-dimensional population activity by incorporating additional classes of latent variables capturing higher-dimensional structure.

show abstract

“…For almost all of the simulations, excluding SGD with label noise, the fraction of active units gradually reduced, long after the loss converged ( For label noise, a slow directed effect was observed but the dynamics were qualitatively different -as predicted by theory [27] and explained in the next section. The fraction of active units did not reduce as much, but the activity of the units did sparsify (Fig S2 [24]. In a sense, once noise is introduced, the network is driven to maximal sparsification in a stochastic manner.…”

Section: Generality Of the Phenomenonmentioning

confidence: 96%

Representational drift as a result of implicit regularization

Ratzon

Derdikman

Barak

2023

Preprint

View full text Add to dashboard Cite

Recent studies show that, even in constant environments, the tuning of single neurons changes over time in a variety of brain regions. This representational drift has been suggested to be a consequence of continuous learning under noise, but its properties are still not fully understood. To uncover the underlying mechanism, we trained an artificial network to perform a predictive coding task. After the loss converged, the activity slowly became sparser. We verified the generality of this phenomenon across modeling choices. This sparseness is a manifestation of drift in the solution space to a flatter area. It is consistent with recent experimental results demonstrating that CA1 spatial code becomes sparser after familiarity. We conclude that learning is divided into three overlapping phases: Fast familiarity with the environment, slow implicit regularization, and a steady state of null drift. These findings open the possibility of inferring learning algorithms from observations of drift statistics.

show abstract

Coordinated drift of receptive fields in Hebbian/anti-Hebbian network models during noisy representation learning

Cited by 22 publications

References 75 publications

Sub-cellular population imaging tools reveal stable apical dendrites in hippocampal area CA3

Sub-cellular population imaging tools reveal stable apical dendrites in hippocampal area CA3

Disentangling Mixed Classes of Covariability in Large-Scale Neural Data

Representational drift as a result of implicit regularization

Contact Info

Product

Resources

About