We inspect and analyze spike recordings of the brain in different situations, such as those resulting from neuro-
related diseases (e.g. seizure, coma), and use this to understand the dynamic characteristics of the neural system
in performing different tasks. Building upon this, we used neuromorphic chips to emulate and predict activity
under different neural parameters and stochasticity. This information is used to construct a phase diagram of brain dynamics.
We combine functions and computational activity and of today’s computers with behavior observed from neurons,
and study its effect on both conventional computing tasks and machine learning tasks. For example, we
modify an algorithm such that it’s activation is similar to neuron activation in the brain to achieve associative
We explore new, coupled computing devices such as magnetic dots and chaotic oscillators for generating dynamic patterns
and manipulate its properties by controlling network parameters. For example, in an annealing-based algorithm,
a system driven chaos has the ability to escape local minima; while a system
driven towards equilibrium can be used to conduct a fine search around the solution space.