Cortical networks with multiple interneuron types generate oscillatory patterns during predictive coding DOI Creative Commons

Kwangjun Lee,

Cyriel M. A. Pennartz, Jorge F. Mejías

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 28, 2024

Abstract Predictive coding (PC) proposes that our brains work as an inference machine, generating internal model of the world and minimizing predictions errors (i.e., differences between external sensory evidence prediction signals). Theoretical models PC often rely on high-level approaches, therefore implementations detailing which neurons or pathways are used to compute adapt representations, well their level agreement with biological circuitry, currently missing. Here we propose a computational PC, integrates neuroanatomically informed hierarchy cortical areas precise laminar organization cell-type-specific connectivity pyramidal, PV, SST VIP cells. Our efficiently performs even in presence noise, by forming latent representations naturalistic visual input (MNIST, fashion-MNIST grayscale CIFAR-10) via Hebbian learning using them predict errors. The assumes both positive negative computed stereotypical pyramidal-PV-SST-VIP circuits same structure but different incoming input. During inference, neural oscillatory activity emerges system due interactions representation error microcircuits, optogenetics-inspired inactivation protocols revealing differentiated role cell types such dynamics. Finally, shows anomalous responses deviant stimuli within series same-image presentations, experimental results mismatch negativity oddball paradigms. We argue constitutes important step better understand mediating real networks. Author summary suggests brain constantly generates expectations about updates these based While this theory is widely accepted, still lack detailed show how specific might carry out processes. Here, present addresses gap including biologically plausible circuitry (pyramidal, SST, cells) connections. It learns form information uses input, adjusting its when occur. found particular play roles processes, oscillations emerge during training also replicates patterns observed experiments where unexpected appear. By integrating anatomical functional details, brings us closer understanding predictive at circuit level.

Language: Английский

Uncertainty-modulated prediction errors in cortical microcircuits DOI Open Access
Katharina A. Wilmes, Mihai A. Petrovici, Shankar Sachidhanandam

et al.

Published: Jan. 22, 2025

Understanding the variability of environment is essential to function in everyday life. The brain must hence take uncertainty into account when updating its internal model world. basis for are prediction errors that arise from a difference between current and new sensory experiences. Although error neurons have been identified layer 2/3 diverse areas, how modulates these learning is, however, unclear. Here, we use normative approach derive should modulate postulate represent uncertainty-modulated (UPE). We further hypothesise circuit calculates UPE through subtractive divisive inhibition by different inhibitory cell types. By implementing calculation UPEs microcircuit model, show types can compute means variances stimulus distribution. With local activity-dependent plasticity rules, computations be learned context-dependently, allow upcoming stimuli their Finally, mechanism enables an organism optimise strategy via adaptive rates.

Language: Английский

Citations

0
Confidence and second-order errors in cortical circuits DOI Creative Commons
Arno Granier, Mihai A. Petrovici, Walter Senn

et al.

PNAS Nexus, Journal Year: 2024, Volume and Issue: 3(9)

Published: Sept. 1, 2024

Abstract Minimization of cortical prediction errors has been considered a key computational goal the cerebral cortex underlying perception, action, and learning. However, it is still unclear how should form use information about uncertainty in this process. Here, we formally derive neural dynamics that minimize under assumption areas must not only predict activity other sensory streams but also jointly project their confidence (inverse expected uncertainty) predictions. In resulting neuronal dynamics, integration bottom-up top-down dynamically modulated based on accordance with Bayesian principle. Moreover, theory predicts existence second-order errors, comparing actual performance. These are propagated through hierarchy alongside classical used to learn weights synapses responsible for formulating confidence. We propose detailed mapping circuitry, discuss entailed functional interpretations, provide potential directions experimental work.

Language: Английский

Citations

2
Uncertainty-modulated prediction errors in cortical microcircuits DOI Open Access
Katharina A. Wilmes, Mihai A. Petrovici, Shankar Sachidhanandam

et al.

Published: Feb. 27, 2024

Understanding the variability of environment is essential to function in everyday life. The brain must hence take uncertainty into account when updating its internal model world. basis for are prediction errors that arise from a difference between current and new sensory experiences. Although error neurons have been identified layer 2/3 diverse areas, how modulates these learning is, however, unclear. Here, we use normative approach derive should modulate postulate represent uncertainty-modulated (UPE). We further hypothesise circuit calculates UPE through subtractive divisive inhibition by different inhibitory cell types. By implementing calculation UPEs microcircuit model, show types can compute means variances stimulus distribution. With local activity-dependent plasticity rules, computations be learned context-dependently, allow upcoming stimuli their Finally, mechanism enables an organism optimise strategy via adaptive rates.

Language: Английский

Citations

0
Uncertainty-modulated prediction errors in cortical microcircuits DOI Open Access
Katharina A. Wilmes, Mihai A. Petrovici, Shankar Sachidhanandam

et al.

Published: Sept. 27, 2024

Understanding the variability of environment is essential to function in everyday life. The brain must hence take uncertainty into account when updating its internal model world. basis for are prediction errors that arise from a difference between current and new sensory experiences. Although error neurons have been identified layer 2/3 diverse areas, how modulates these learning is, however, unclear. Here, we use normative approach derive should modulate postulate represent uncertainty-modulated (UPE). We further hypothesise circuit calculates UPE through subtractive divisive inhibition by different inhibitory cell types. By implementing calculation UPEs microcircuit model, show types can compute means variances stimulus distribution. With local activity-dependent plasticity rules, computations be learned context-dependently, allow upcoming stimuli their Finally, mechanism enables an organism optimise strategy via adaptive rates.

Language: Английский

Citations

0
Cortical networks with multiple interneuron types generate oscillatory patterns during predictive coding DOI Creative Commons

Kwangjun Lee,

Cyriel M. A. Pennartz, Jorge F. Mejías

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 28, 2024

Abstract Predictive coding (PC) proposes that our brains work as an inference machine, generating internal model of the world and minimizing predictions errors (i.e., differences between external sensory evidence prediction signals). Theoretical models PC often rely on high-level approaches, therefore implementations detailing which neurons or pathways are used to compute adapt representations, well their level agreement with biological circuitry, currently missing. Here we propose a computational PC, integrates neuroanatomically informed hierarchy cortical areas precise laminar organization cell-type-specific connectivity pyramidal, PV, SST VIP cells. Our efficiently performs even in presence noise, by forming latent representations naturalistic visual input (MNIST, fashion-MNIST grayscale CIFAR-10) via Hebbian learning using them predict errors. The assumes both positive negative computed stereotypical pyramidal-PV-SST-VIP circuits same structure but different incoming input. During inference, neural oscillatory activity emerges system due interactions representation error microcircuits, optogenetics-inspired inactivation protocols revealing differentiated role cell types such dynamics. Finally, shows anomalous responses deviant stimuli within series same-image presentations, experimental results mismatch negativity oddball paradigms. We argue constitutes important step better understand mediating real networks. Author summary suggests brain constantly generates expectations about updates these based While this theory is widely accepted, still lack detailed show how specific might carry out processes. Here, present addresses gap including biologically plausible circuitry (pyramidal, SST, cells) connections. It learns form information uses input, adjusting its when occur. found particular play roles processes, oscillations emerge during training also replicates patterns observed experiments where unexpected appear. By integrating anatomical functional details, brings us closer understanding predictive at circuit level.

Language: Английский

Citations

0