A carregar...
Publicação
Identifying Invariant Representations underlying Neural Computation
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 5.17 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
"One of the fundamental goals of systems neuroscience is understanding
how the coordinated activity of populations of neurons gives rise to flexible
behavior. The language of dynamical systems has been invaluable for
modeling and describing the neural computation underlying various cognitive
behaviors through the spatiotemporal evolution of the neural state.
Despite the heterogeneity of neural activity, the dynamical primitives in
biological and artificial neural networks engaged in the same behavior
demonstrate striking similarity. In this thesis, we propose novel machine
learning techniques that facilitate modeling and analysis across diverse
recording sessions, subjects, tasks and experimental conditions within a
unified dynamical space. Specifically, we develop a hierarchical framework
for learning a family of dynamical systems from diverse datasets
that can concisely capture the similarity and diversity underlying neural
recordings.(...)"
Descrição
Palavras-chave
dynamical systems neural recordings neural computation modeling dynamical systems