Dataset Information

Revealing the information contents of memory within the stimulus information representation framework.

ABSTRACT: The information contents of memory are the cornerstone of the most influential models in cognition. To illustrate, consider that in predictive coding, a prediction implies that specific information is propagated down from memory through the visual hierarchy. Likewise, recognizing the input implies that sequentially accrued sensory evidence is successfully matched with memorized information (categorical knowledge). Although the existing models of prediction, memory, sensory representation and categorical decision are all implicitly cast within an information processing framework, it remains a challenge to precisely specify what this information is, and therefore where, when and how the architecture of the brain dynamically processes it to produce behaviour. Here, we review a framework that addresses these challenges for the studies of perception and categorization-stimulus information representation (SIR). We illustrate how SIR can reverse engineer the information contents of memory from behavioural and brain measures in the context of specific cognitive tasks that involve memory. We discuss two specific lessons from this approach that generally apply to memory studies: the importance of task, to constrain what the brain does, and of stimulus variations, to identify the specific information contents that are memorized, predicted, recalled and replayed. This article is part of the Theo Murphy meeting issue 'Memory reactivation: replaying events past, present and future'.

SUBMITTER: Schyns PG

PROVIDER: S-EPMC7209912 | biostudies-literature | 2020 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Revealing the information contents of memory within the stimulus information representation framework.

Schyns Philippe G PG Zhan Jiayu J Jack Rachael E RE Ince Robin A A RAA

Philosophical transactions of the Royal Society of London. Series B, Biological sciences 20200406 1799

The information contents of memory are the cornerstone of the most influential models in cognition. To illustrate, consider that in predictive coding, a prediction implies that specific information is propagated down from memory through the visual hierarchy. Likewise, recognizing the input implies that sequentially accrued sensory evidence is successfully matched with memorized information (categorical knowledge). Although the existing models of prediction, memory, sensory representation and cat ...[more]

PMID: 32248774

Similar Datasets

Project description:Machine Learning models have been frequently used in transcriptome analyses. Particularly, Representation Learning (RL), e.g., autoencoders, are effective in learning critical representations in noisy data. However, learned representations, e.g., the "latent variables" in an autoencoder, are difficult to interpret, not to mention prioritizing essential genes for functional follow-up. In contrast, in traditional analyses, one may identify important genes such as Differentially Expressed (DiffEx), Differentially Co-Expressed (DiffCoEx), and Hub genes. Intuitively, the complex gene-gene interactions may be beyond the capture of marginal effects (DiffEx) or correlations (DiffCoEx and Hub), indicating the need of powerful RL models. However, the lack of interpretability and individual target genes is an obstacle for RL's broad use in practice. To facilitate interpretable analysis and gene-identification using RL, we propose "Critical genes", defined as genes that contribute highly to learned representations (e.g., latent variables in an autoencoder). As a proof-of-concept, supported by eXplainable Artificial Intelligence (XAI), we implemented eXplainable Autoencoder for Critical genes (XA4C) that quantifies each gene's contribution to latent variables, based on which Critical genes are prioritized. Applying XA4C to gene expression data in six cancers showed that Critical genes capture essential pathways underlying cancers. Remarkably, Critical genes has little overlap with Hub or DiffEx genes, however, has a higher enrichment in a comprehensive disease gene database (DisGeNET) and a cancer-specific database (COSMIC), evidencing its potential to disclose massive unknown biology. As an example, we discovered five Critical genes sitting in the center of Lysine degradation (hsa00310) pathway, displaying distinct interaction patterns in tumor and normal tissues. In conclusion, XA4C facilitates explainable analysis using RL and Critical genes discovered by explainable RL empowers the study of complex interactions.

Dataset Information

Revealing the information contents of memory within the stimulus information representation framework.

Publications

Revealing the information contents of memory within the stimulus information representation framework.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets