Representational learning by optimization of neural manifolds in an olfactory memory network
12 affiliations
- 1Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland
- 2Institute of Neuroscience, CAS Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai, China
- 3University of Basel, Basel, Switzerland
- 4Center for Computational Neuroscience, Flatiron Institute, New York, NY, USA
- 5Physics Department, Center for Brain Science, and Kempner Institute, Harvard University, Cambridge, MA, USA
- 6Brain Research Institute, University of Zürich, Zurich, Switzerland
- 7Neuroscience Center Zurich and Brain Research Institute, University of Zurich, Zurich, Switzerland
- 8Neuroscience Center Zürich, Zurich, Switzerland
- 9Sorbonne Université, CNRS, Inserm, Center of Neuroscience NeuroSU, Paris, France
- 10Center for Neural Science, New York University, New York, NY, USA
- 11Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland. [email protected]
- 12University of Basel, Basel, Switzerland. [email protected]
Abstract
Cognitive brain functions rely on experience-dependent internal representations of relevant information. Such representations are organized by attractor dynamics or other mechanisms that constrain population activity onto “neural manifolds”. Quantitative analyses of representational manifolds are complicated by their potentially complex geometry, particularly in the absence of attractor states. Here we trained juvenile and adult zebrafish in an odor discrimination task and measured neuronal population activity to analyze representations of behaviorally relevant odors in telencephalic area pDp, the homolog of piriform cortex. No obvious signatures of attractor dynamics were detected. However, olfactory discrimination training selectively enhanced the separation of neural manifolds representing task-relevant odors from other representations, consistent with predictions of autoassociative network models endowed with precise synaptic balance. Analytical approaches using the framework of manifold capacity revealed multiple geometrical modifications of representational manifolds that supported the classification of task-relevant sensory information. Manifold capacity predicted odor discrimination across individuals better than other descriptors of population activity, indicating a close link between manifold geometry and behavior. Hence, pDp and possibly related recurrent networks store information in the geometry of representational manifolds, resulting in joint sensory and semantic maps that may support distributed learning processes.
Related
- Decomposing deep neural network-brain representational similarity reveals distinct sources across the human visual hierarchy
- Multidimensional semantic representations emerge from multi-frequency representational similarity learning
- Case Report: Representational geometry of multi-digit coordination in the macaque grasping network for intracortical BCIs
- Continual-learning rules shape representational drift
- Representational geometry of punishment learning across basolateral amygdala neurons
BibTeX
@article{hu2026representational,
title = {{Representational learning by optimization of neural manifolds in an olfactory memory network}},
author = {Bo Hu and Nesibe Z. Temiz and Chi-Ning Chou and Peter Rupprecht and Claire Meissner-Bernard and Benjamin Titze and SueYeon Chung and Rainer W. Friedrich},
journal = {Nature Neuroscience},
year = {2026},
month = sep,
doi = {10.1038/s41593-026-02429-3},
url = {https://doi.org/10.1038/s41593-026-02429-3},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Representational learning by optimization of neural manifolds in an olfactory memory network
AU - Bo Hu
AU - Nesibe Z. Temiz
AU - Chi-Ning Chou
AU - Peter Rupprecht
AU - Claire Meissner-Bernard
AU - Benjamin Titze
AU - SueYeon Chung
AU - Rainer W. Friedrich
PY - 2026
DA - 2026/09/10
JO - Nature Neuroscience
DO - 10.1038/s41593-026-02429-3
UR - https://doi.org/10.1038/s41593-026-02429-3
AB - Cognitive brain functions rely on experience-dependent internal representations of relevant information. Such representations are organized by attractor dynamics or other mechanisms that constrain population activity onto “neural manifolds”. Quantitative analyses of representational manifolds are complicated by their potentially complex geometry, particularly in the absence of attractor states. Here we trained juvenile and adult zebrafish in an odor discrimination task and measured neuronal population activity to analyze representations of behaviorally relevant odors in telencephalic area pDp, the homolog of piriform cortex. No obvious signatures of attractor dynamics were detected. However, olfactory discrimination training selectively enhanced the separation of neural manifolds representing task-relevant odors from other representations, consistent with predictions of autoassociative network models endowed with precise synaptic balance. Analytical approaches using the framework of manifold capacity revealed multiple geometrical modifications of representational manifolds that supported the classification of task-relevant sensory information. Manifold capacity predicted odor discrimination across individuals better than other descriptors of population activity, indicating a close link between manifold geometry and behavior. Hence, pDp and possibly related recurrent networks store information in the geometry of representational manifolds, resulting in joint sensory and semantic maps that may support distributed learning processes.
SN - 1097-6256
SN - 1546-1726
ER -