Spiking network model of the cerebellum as a reinforcement learning machine DOI Creative Commons

Rin Kuriyama,

Hideyuki Yoshimura,

Tadashi Yamazaki

et al.

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

Published: June 28, 2024

The cerebellum has been considered to perform error-based supervised learning via long-term depression (LTD) at synapses between parallel fibers and Purkinje cells (PCs). Since the discovery of multiple synaptic plasticity other than LTD, recent studies have suggested that synergistic mechanisms could enhance capability cerebellum. Indeed, we proposed a concept cerebellar as reinforcement (RL) machine. However, there is still gap conceptual algorithm its detailed implementation. To close this gap, in research, implemented spiking network an RL model continuous time space, based on known anatomical properties We confirmed our successfully learned state value solved mountain car task, simple benchmark. Furthermore, demonstrated ability solve delay eyeblink conditioning task using biologically plausible internal dynamics. Our research provides solid foundation for theory challenges classical view primarily

Language: Английский

Human Purkinje cells outperform mouse Purkinje cells in dendritic complexity and computational capacity DOI Creative Commons
Stefano Masoli, Diana Sánchez-Ponce, Nora Vrieler

et al.

Communications Biology, Journal Year: 2024, Volume and Issue: 7(1)

Published: Jan. 2, 2024

Abstract Purkinje cells in the cerebellum are among largest neurons brain and have been extensively investigated rodents. However, their morphological physiological properties remain poorly understood humans. In this study, we utilized high-resolution reconstructions unique electrophysiological recordings of human ex vivo to generate computational models estimate capacity. An inter-species comparison showed that cell had similar fractal structures but were larger than those mouse cells. Consequently, given a spine density (2/μm), hosted approximately 7.5 times more dendritic spines mice. Moreover, higher complexity usually emitted 2–3 main trunks instead one. Intrinsic electro-responsiveness was between two species, model simulations revealed dendrites could process ~6.5 (n = 51 vs. n 8) input patterns Thus, while maintained spike discharge rodents during evolution, they developed complex dendrites, enhancing

Language: Английский

Citations

15
Cerebellar control of targeted tongue movements DOI Creative Commons
Lorenzo Bina,

Camilla Ciapponi,

Si‐yang Yu

et al.

The Journal of Physiology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 26, 2025

Abstract The cerebellum is critical for coordinating movements related to eating, drinking and swallowing, all of which require proper control the tongue. Cerebellar Purkinje cells can encode tongue movements, but it unclear how their simple spikes complex induce changes in shape that contribute goal‐directed movements. To study these relations, we recorded stimulated vermis hemispheres mice during spontaneous licking from a stationary or moving water spout. We found rhythmic with both spikes. Increased spike firing protrusion induces ipsiversive bending Unexpected target location trigger alter subsequent licks, adjusting trajectory. Furthermore, observed increased behavioural state at start end bouts. Using machine learning, confirmed alterations cell activity accompany licking, different often exerting heterogeneous encoding schemes. Our data highlight directional movement paramount cerebellar function modulation are complementary acquisition execution sensorimotor coordination. These results bring us closer understanding clinical implications disorders swallowing. image Key points When drinking, make directed towards source. fire rhythmically tune position source affects direction report also adjust right direction.

Language: Английский

Citations

1
Linking cellular-level phenomena to brain architecture: the case of spiking cerebellar controllers DOI Creative Commons
Egidio D’Angelo, Alberto Antonietti, Alice Geminiani

et al.

Neural Networks, Journal Year: 2025, Volume and Issue: 188, P. 107538 - 107538

Published: April 23, 2025

Linking cellular-level phenomena to brain architecture and behavior is a holy grail for theoretical computational neuroscience. Advances in neuroinformatics have recently allowed scientists embed spiking neural networks of the cerebellum with realistic neuron models multiple synaptic plasticity rules into sensorimotor controllers. By minimizing distance (error) between desired actual sensory state, exploiting prediction, cerebellar network acquires knowledge about body-environment interaction generates corrective signals. In doing so, implements generalized algorithm, allowing it "to learn predict timing correlated events" rich set behavioral contexts. Plastic changes evolve trial by are distributed over synapses, regulating neuronal discharge fine-tuning high-speed movements on millisecond timescale. Thus, built-in controllers, among various approaches studying function, helping reveal substrates learning signal coding, opening new frontiers predictive computing autonomous robots.

Language: Английский

Citations

0
Cerebellar contribution to multisensory integration: A computational modeling exploration DOI Creative Commons
Riccardo Cavadini, Luca Casartelli, Alessandra Pedrocchi

et al.

APL Bioengineering, Journal Year: 2025, Volume and Issue: 9(2)

Published: April 24, 2025

The remarkable ability of the human brain to create a coherent perception reality relies heavily on multisensory integration—the complex process combining inputs from different senses. While this mechanism is fundamental our understanding world, its underlying neural architecture remains partially unknown. This study investigates role cerebellum in integration through novel computational approach inspired by clinical observations patient with cerebellar agenesis. With reference data comparing an acerebellar age-matched control subjects, we exploited biologically realistic spiking networks model both conditions. Our framework enables testing multiple network configurations and parameters, effectively replicating extending experiments silico. To enhance accessibility promote broader adoption among researchers, complemented user-friendly web-based interface, eliminating need for programming expertise. results closely mirror findings, providing support critical contribution integration. Beyond being consistent proof concept previous observations, introduces versatile platform models newly developed interface. Thus, work not only advances sensory processing but also establishes robust methodology future investigations mechanisms.

Language: Английский

Citations

0
Purkinje cell models: past, present and future DOI Creative Commons

Elías Mateo Fernández Santoro,

Arun Karim,

Pascal Warnaar

et al.

Frontiers in Computational Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: July 10, 2024

The investigation of the dynamics Purkinje cell (PC) activity is crucial to unravel role cerebellum in motor control, learning and cognitive processes. Within cerebellar cortex (CC), these neurons receive all incoming sensory information, transform it generate entire output. relatively homogenous repetitive structure CC, common vertebrate species, suggests a single computation mechanism shared across PCs. While PC models have been developed since 70′s, comprehensive review contemporary currently lacking. Here, we provide an overview models, ranging from ones focused on intracellular dynamics, through complex which include synaptic extrasynaptic inputs. We how can reproduce physiological neuron, including firing patterns, current multistable plateau potentials, calcium signaling, intrinsic plasticity input/output computations. consider focusing both somatic dendritic Our provides critical performance analysis with respect known data. expect our synthesis be useful guiding future development computational that capture real-life context

Language: Английский

Citations

1
Spiking network model of the cerebellum as a reinforcement learning machine DOI Creative Commons

Rin Kuriyama,

Hideyuki Yoshimura,

Tadashi Yamazaki

et al.

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

Published: June 28, 2024

The cerebellum has been considered to perform error-based supervised learning via long-term depression (LTD) at synapses between parallel fibers and Purkinje cells (PCs). Since the discovery of multiple synaptic plasticity other than LTD, recent studies have suggested that synergistic mechanisms could enhance capability cerebellum. Indeed, we proposed a concept cerebellar as reinforcement (RL) machine. However, there is still gap conceptual algorithm its detailed implementation. To close this gap, in research, implemented spiking network an RL model continuous time space, based on known anatomical properties We confirmed our successfully learned state value solved mountain car task, simple benchmark. Furthermore, demonstrated ability solve delay eyeblink conditioning task using biologically plausible internal dynamics. Our research provides solid foundation for theory challenges classical view primarily

Language: Английский

Citations

0