Apprendimento Profondo e Neuroscienze: un'integrazione

The thesis regards the interconnection between neuroscience and computer technology. Efforts towards research in artificial intelligence (AI) have reached unprecedented levels during the last years, this progress came mainly from recent advancements from the deep learning sub-field of research. These algorithms take inspirations and mimic features present in biological brains by the virtue of their brain-like computation. A particular type of deep neural networks, called convolutional neural networks, possess features which are directly inspired by neuroscientific notions of simple cells and complex cells in visual neuroscience and his structure is reminiscent of the the visual cortex ventral pathway. Neuroscientists look to artificial neural networks as a research tool for the interpretation of neurobiological phenomena, to help them to reveal the underlying processes of the brain, and engineers look to neurobiology for new ideas to solve problems more complex than those based on conventional design techniques. Although history has shown the ups and downs about interest in conjugate neuroscience and artificial intelligence, nowadays this gap in increasingly thinning. This growing importance of deep neural networks in neuroscience grants synergic work to shed light upon how the brain works, and neuroscientists seek to implement in their research work these neural networks as an effective tool, together with recent and state-of-the-art neuroimaging tools. This thesis introduces to the reader the key components of neural networks and illustrate the application of deep learning algorithms in neuroscience. Namely, a compendium of neural networks features, an overview on one of the most used deep neural networks, the convolutional neural networks, and a bibliographical survey about the implementation of deep neural networks in neuroscience, starting from the similarities between these networks and the human vision system and progressing through the implementation of aspects like visual attention and neuroimaging.