Biology subjects

Jaras, I.

Publications and source records attributed to Jaras, I..

3 recordsLinked to original sources

bioRxiv2025

"Backpropagation and the brain" realized in cortical error neuron microcircuits

Neural responses to mismatches between expected and actual stimuli have been widely reported across different species. How does the brain use such error signals for learning? While global error signals can be useful, their ability to learn complex computation at the scale observed in the brain is lacking. In comparison, more local, neuron-specific error signals enable superior performance, but their computation and propagation remain unclear. Motivated by the breakthrough of deep learning, this has inspired the backpropagation and the brain hypothesis, i.e. that the brain implements a form of the error backpropagation algorithm. In this work, we introduce a biologically motivated, multi-area cortical microcircuit model, implementing error backpropagation under consideration of recent physiological evidence. We model populations of cortical pyramidal cells acting as representation and error neurons, with bio-plausible local and inter-area connectivity, guided by experimental observations of connectivity of the primate visual cortex. In our model, all information transfer is biologically motivated, inference and learning occur without phases, and network dynamics demonstrably approximate those of error backpropagation. We show the capabilities of our model on a wide range of benchmarks, and compare to other models, such as dendritic hierarchical predictive coding. In particular, our model addresses shortcomings of other theories in terms of scalability to many cortical areas. Finally, we make concrete predictions, which differentiate it from other theories, and which can be tested in experiment.

neuroscience↗

bioRxiv2023

Unveiling the role of local metabolic constraints on the structure and activity of spiking neural networks

Understanding the intricate interplay between neural dynamics and metabolic constraints is crucial for unraveling the mysteries of the brain. Despite the significance of this relationship, specific details concerning the impact of metabolism on neuronal dynamics and neural network architecture remain elusive, creating a notable gap in the existing literature. This study employs an energy-dependent neuron and plasticity model to analyze the role of local metabolic constraints in shaping both the dynamics and structure of Spiking Neural Networks (SNN). Specifically, an energy-dependent version of the leaky integrate-and-fire model is utilized, along with a three-factor learning rule that incorporates postsynaptic available energy as the third factor. These models allow for fine-tuning sensitivity in the presence of energy imbalances. Analytical expressions predicting the networks activity and structure are derived, and a fixed-point analysis reveals the emergence of attractor states characterized by neuronal and synaptic sensitivity to energy imbalances. Analytical findings are validated through numerical simulations using an excitatory-inhibitory network. Furthermore, these simulations enable the study of SNN activity and structure under conditions simulating metabolic impairment. In conclusion, by employing energy-dependent models with adjustable sensitivity to energy imbalances, our study advances the understanding of how metabolic constraints shape SNN dynamics and structure. Moreover, in light of compelling evidence linking neuronal metabolic impairment to neurodegenerative diseases, the incorporation of local metabolic constraints into the investigation of neuronal network structure and activity opens an intriguing avenue for inspiring the development of therapeutic interventions.

neuroscience↗

bioRxiv2020

Extending the integrate-and-fire model to account for metabolic dependencies

It is widely accepted that the brain, like any other physical system, is subjected to physical constraints restricting its operation. The brains metabolic demands are particularly critical for proper neuronal function, but the impact of these constraints is still poorly understood. Detailed single-neuron models are recently integrating metabolic constraints, but the computational resources these models need, make it difficult to explore the dynamics of extended neural networks imposed by such constraints. Thus, there is a need for a simple-enough neuron model that incorporates metabolic activity and allows us to explore neural network dynamics. This work introduces an energy-dependent leaky integrate-and-fire (LIF) neuronal model extension to account for the effects of metabolic constraints on the single-neuron behavior (EDLIF). This simple energy-dependent model shows better performance predicting real spikes trains -in spike coincidence measure sense-than the classical leaky integrate-and-fire model. It can describe the relationship between the average firing rate and the ATP cost, and replicate a neurons behavior under a clinical setting such as amyotrophic lateral sclerosis. The simplicity of the energy-dependent model presented here, makes it computationally efficient and thus, suitable to study the dynamics of large neural networks. Author summaryAny physical system or biological tissue is restricted by physical constraints bounding their behavior, and the brain is not free from these constraints. Energetic disorders in the brain have been linked to several neurodegenerative diseases, highlighting the relevance of maintaining a critical balance between energy production and consumption in neurons. These observations motivate the development of mathematical tools that can help to understand the dependence of the brains behavior in metabolism. One of the essential building blocks to achieve this task is the mathematical representation of neurons through models, allowing computational simulations of single-neurons and neural networks. Here we construct a simple and computational cheap energy-dependent neuron model that allows the study of neurons behavior under an energetic perspective. The introduced neuron model is contrasted with one of the widest-used neuron models and shows better prediction capabilities when real neuron recordings are used. Our model is suitable for replicating neurons behavior under a specific neurodegenerative disease, which cannot be achieved by the abovementioned popular model. Our simple model is promising because it allows the simulation and study of neuronal networks under a metabolic-dependent perspective.

neuroscience↗