Neural Dynamics and Machine Intelligence
How does intelligent behavior sprout out of brain tissue? We study how computation emerges from large networks of neurons and search for principled dynamics that lead to robust behavior. We apply bottom-up and top-bottom approaches: model neural activity to explain experimental observations and derive fundamental theories of neural computation.
Modern Machine Learning techniques give us artificial neural networks that present, or at least mimic, a form of intelligence. These systems bring new opportunities and new challenges to neuroscience. Here we have far simpler systems than the brain, which we design and build, yet we do not fully understand how and why they work so well. Did we unlock some of the mysteries of the mind? Are we on the right track towards a mechanistic theory for cognition, or are we climbing the highest tree in our forest, hoping to reach the moon? We combine machine learning practices with tools from statistical mechanics, dynamical systems, and information theory. We study what makes artificial networks tick and then look at them with a critical eye, searching for biological relevance.
