Dataset Information

The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.

ABSTRACT: The curse of dimensionality plagues models of reinforcement learning and decision making. The process of abstraction solves this by constructing variables describing features shared by different instances, reducing dimensionality and enabling generalization in novel situations. Here, we characterized neural representations in monkeys performing a task described by different hidden and explicit variables. Abstraction was defined operationally using the generalization performance of neural decoders across task conditions not used for training, which requires a particular geometry of neural representations. Neural ensembles in prefrontal cortex, hippocampus, and simulated neural networks simultaneously represented multiple variables in a geometry reflecting abstraction but that still allowed a linear classifier to decode a large number of other variables (high shattering dimensionality). Furthermore, this geometry changed in relation to task events and performance. These findings elucidate how the brain and artificial systems represent variables in an abstract format while preserving the advantages conferred by high shattering dimensionality.

SUBMITTER: Bernardi S

PROVIDER: S-EPMC8451959 | biostudies-literature | 2020 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.

Bernardi Silvia S Benna Marcus K MK Rigotti Mattia M Munuera Jérôme J Fusi Stefano S Salzman C Daniel CD

Cell 20201014 4

The curse of dimensionality plagues models of reinforcement learning and decision making. The process of abstraction solves this by constructing variables describing features shared by different instances, reducing dimensionality and enabling generalization in novel situations. Here, we characterized neural representations in monkeys performing a task described by different hidden and explicit variables. Abstraction was defined operationally using the generalization performance of neural decoder ...[more]

PMID: 33058757

Similar Datasets

Project description:Suicide is one of the leading causes of death globally for all ages, and as such presents a very serious problem for clinicians worldwide. However, the underlying neurobiological pathology remains to a large extent unknown. In order to address this gap, we have carried out a genome-wide investigation of the gene expression in the amygdala, hippocampus, prefrontal cortex and thalamus in post-mortem brain samples obtained from 20 suicide completers and 7 control subjects. By KEGG enrichment analysis indicated we identified novel clusters of downregulated pathways involved in antigen neutralization and autoimmune thyroid disease (amygdala, thalamus), decreased axonal plasticity in the hippocampus. Two upregulated pathways were involved in neuronal death in the hippocampus and olfactory transduction in the thalamus and the prefrontal cortex. Autoimmune thyroid disease pathway was downregulated only in females. Metabolic pathways involved in Notch signaling amino acid metabolism and unsaturated lipid synthesis were thalamus-specific. Suicide-associated changes in the expression of several genes and pseudogenes that point to various functional mechanisms possibly implicated in the pathology of suicide. Two genes (SNORA13 and RNU4-2) involved in RNA processing were common to all brain regions analyzed. Most of the identified gene expression changes were related to region-specific dysregulated manifestation of genetic and epigenetic mechanisms underlying neurodevelopmental disorders (SNORD114-10, SUSd1), motivation, addiction and motor disorders (CHRNA6), long-term depression (RAB3B), stress response, major depression and schizophrenia (GFAP), signal transduction at the neurovascular unit (NEXN) and inhibitory neurotransmission in spatial learning, neural plasticity (CALB2; CLIC6, ENPP1). Some of the differentially expressed genes were brain specific non-coding RNAs involved in the regulation of translation (SNORA13). One, (PARM1) is a potential oncogene and prognostic biomarker for colorectal cancer with no known function in the brain. Disturbed gene expression involved in antigen neutralization, autoimmunity, neural plasticity, stress response, signal transduction at the neurovascular unit, dysregulated nuclear RNA processing and translation and epigenetic imprinting signatures is associated with suicide and point to regulatory non-coding RNAs as potential targets of new drugs development.

Dataset Information

The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.

Publications

The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets