Video Experimental Relacionado
Updated: Feb 13, 2026

03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.1K
Plasticidad de Múltiples Escalas Controlada por Astrocitos para Aprendizaje Continuo en Línea en Redes Neuronales de
1School of Computer and Data Science, Minjiang University, Fuzhou, China.
Frontiers in neuroscience
|February 12, 2026
Resumen
Este estudio presenta la Plasticidad de Múltiples Escalas Controlada por Astrocitos (AGMP), un nuevo marco de aprendizaje en línea para Redes Neuronales de Espigas (SNN). AGMP permite un aprendizaje continuo robusto, superando el olvido catastrófico y las limitaciones de memoria en SNN.
Área de la Ciencia:
- Ingeniería Neuromórfica
- Neurociencia Computacional
- Inteligencia Artificial
Sus antecedentes:
- Las Redes Neuronales de Espigas (SNN) ofrecen una computación eficiente energéticamente y basada en eventos, ideal para datos sensoriales en tiempo real.
- El entrenamiento en línea y continuo de SNN profundas enfrenta desafíos como cuellos de botella de memoria con Backpropagation-through-Time (BPTT) y el dilema estabilidad-plasticidad en las reglas de aprendizaje local.
Objetivo del estudio:
- Desarrollar un marco de aprendizaje en línea escalable para SNN que aborde las limitaciones de los métodos de entrenamiento existentes.
- Introducir un mecanismo inspirado biológicamente, la Plasticidad de Múltiples Escalas Controlada por Astrocitos (AGMP), para un aprendizaje continuo robusto en SNN.
Principales métodos:
- AGMP aumenta las trazas de elegibilidad con una señal de enseñanza de difusión y un mecanismo de control mediado por astrocitos.
- Una variable astrocítica lenta modula dinámicamente la plasticidad en función de la actividad neuronal, suprimiendo las actualizaciones durante períodos estables y permitiendo la adaptación durante los cambios de distribución.
- El marco se evaluó en puntos de referencia neuromórficos (N-Caltech101, DVS128 Gesture, SHD) y tareas de Aprendizaje Continuo de Clase Incremental (Split CIFAR-100).
Principales resultados:
- AGMP logra una precisión comparable a BPTT fuera de línea manteniendo una complejidad de memoria temporal constante O(1).
- En el Aprendizaje Continuo de Clase Incremental, AGMP reduce significativamente el olvido catastrófico sin necesidad de búferes de repetición.
- AGMP supera las reglas de aprendizaje en línea existentes de última generación en escenarios de aprendizaje continuo.
Conclusiones:
- AGMP presenta un enfoque biológicamente fundamentado y amigable con el hardware para el aprendizaje permanente en agentes autónomos.
- El método propuesto ofrece una solución viable para el entrenamiento en línea y continuo robusto y eficiente de SNN.
Videos de Conceptos Relacionados
Plasticity
3.1K
Plasticity is the property where an object loses its elasticity and undergoes irreversible deformation, even after the deformation forces are eliminated. If a material deforms irreversibly without increasing stress or load, then this is called ideal plasticity. For example, when a force is applied to an aluminum rod, it changes its shape, but it does not return to its original shape once the force is removed. Plastic deformation or ductility is thus a permanent deformation or change in the...
3.1K
Plasticizers
378
Water-reducers, or plasticizers, are chemical admixtures used in concrete to improve strength and workability. These additives reduce the water-cement ratio without compromising workability, lower the cement content while maintaining the same workability, or increase workability to assist concrete placement in inaccessible areas.
Plasticizers function by using surface-active agents to create repulsive electrostatic forces between cement particles. This dispersion enhances the concrete's...
Plasticizers function by using surface-active agents to create repulsive electrostatic forces between cement particles. This dispersion enhances the concrete's...
378
Network Covalent Solids
16.2K
Network covalent solids contain a three-dimensional network of covalently bonded atoms as found in the crystal structures of nonmetals like diamond, graphite, silicon, and some covalent compounds, such as silicon dioxide (sand) and silicon carbide (carborundum, the abrasive on sandpaper). Many minerals have networks of covalent bonds.
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
16.2K
Protein Networks
4.6K
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
4.6K
Plastic Behavior
585
A material's elastic behavior is characterized by the disappearance of stress once the load is removed, allowing the material to return to its original state. However, when stress surpasses the yield point, yielding commences, marking the onset of plastic deformation or permanent set. This change from elastic to plastic behavior is influenced by the peak stress value and the duration before the load is removed. An intriguing observation occurs when a specimen is loaded, unloaded, and...
585
Plastic Deformations
477
It is essential to understand how structural members behave under plastic deformation when the bending stress exceeds the material's yield strength. This state of deformation permanently alters the shape of the member, in contrast to the linear elastic behavior observed before yielding. The strain at any point in the member is expressed in terms of maximum strain. Notably, the neutral axis, which coincides with the centroid during elastic bending, shifts away from the centroid under plastic...
477

