Efeito da plasticidade na sincronização em redes de neurônios com burst
Abstract
Resumo: O cerebro pode ser pensado como uma rede de neuronios, cuja estrutura complexa possibilita a grande diversidade de funcoes que ele exibe. Essa estrutura cerebral pode se alterar com o tempo, em um processo chamado plasticidade neuronal. Aprendizagem e formacao de memoria sao exemplos de fenomenos que dependem da plasticidade. Algumas regras de plasticidade ja foram propostas, entre elas a plasticidade dependente do tempo de rajada (BTDP), que leva em conta o tempo relativo entre o inicio de rajadas de dois neuronios conectados. A sincronizacao de neuronios tambem e um fenomeno encontrado no sistema nervoso, e esta relacionada a diversas funcoes cerebrais e tambem a doencas. Sabe-se que a plasticidade pode alterar o nivel de sincronizacao neuronal. Modelos matematicos de redes neuronais tem sido utilizados para investigar os comportamentos encontrados no sistema nervoso. Nesta dissertacao, construimos uma rede neuronal com neuronios modelados pelo modelo de Rulkov, e conectados atraves de sinapses quimicas excitatorias. Os pesos sinapticos sao alterados no tempo atraves da plasticidade BTDP. Duas arquiteturas de conexao sao investigadas: aleatoria e de mundo pequeno. Buscamos estudar como a plasticidade BTDP altera a estrutura e a sincronizacao da rede. Foi observado que, dependendo do peso sinaptico inicial, toda ou parte da rede pode ter as sinapses fortalecidas ao maior peso sinaptico possivel. Para a rede aleatoria, dois mecanismos para esse fortalecimento foram identificados, um devido a uma sincronizacao previamente presente, e outro devido a impossibilidade de grandes valores de defasagem de rajadas para neuronios de frequencia alta. A coexistencia desses mecanismos tambem foi verificada. Nesse caso a rede pode apresentar uma segregacao em dois aglomerados apos a evolucao, desde que o nivel de ruido na rede seja baixo o suficiente. Verificou-se tambem que, em quase todos os casos, a sincronizacao e aumentada devido a plasticidade, seguindo o fortalecimento das sinapses da rede. Para a rede de mundo pequeno, foi verificado que o estado final da rede pode apresentar propriedades locais ou nao-locais, dependendo do peso sinaptico inicial das conexoes. Palavras-chaves: redes neuronais. neuronios de Rulkov. plasticidade. sincronizacao Abstract: The brain can be thought of as a network of neurons, whose complex structure allows the brain to perform the great diversity of functions that it is known for. This structure can be altered with time, in a process called neuroplasticity. Learning and memory formation are some phenomena that depend on brain plasticity. Some plasticity rules have been proposed to explain its behavior, such as the burst-timing-dependent plasticity (BTDP). This plasticity rule considers the relative time of the burst start of two connected neurons to modify the brain structure. Synchronization of neurons is another phenomenon found on the nervous system, and is related to many brain functions and nervous disorders. It is known that plasticity can alter the neuronal synchronization level. Mathematical models of neuronal networks have been used to investigate behaviors found on the nervous system. In this work, we build a neuronal network with neurons modeled by the Rulkov model, and connected via excitatory chemical synapses. The synaptic weights are modified following the BTDP rule. Two network architectures are studied: random and small-world network. We aim to understand how BTDP modifies the structure and synchronization of the network. It was observed that, depending on the initial synaptic weight, the whole network or part of it can have its synapses strengthened to the maximum possible synaptic weight. For the random network, two mechanisms for this strengthening were identified. The first is due to a preexisting synchronization and the second due to the impossibility of burst latencies assuming high values for high-frequency neurons. The coexistence of the two mechanisms is also verified. In this case, the network can exhibit a segregation in two clusters on the asymptotic state, but only if the noise level on the network is low enough. It was also verified that, for almost all cases, the network synchronization level is increased due to the plasticity, following the strengthening of network synapses. For the small-world network, it was verified that the final state of the network can exhibit local or non-local properties, depending on the initial synaptic weight of the connections. Key-words: neuronal networks. Rulkov neurons. plasticity. synchronization
Collections
- Dissertações [205]