Aims. Find mechanisms of information processing in intelligent agents (animals, humans, robots, software systems). Mechanisms should be robust against noise, cheap and simple and data-efficient.
Premise. Intelligence results from the interaction of many dynamical-computational mechanisms of which only a few have been identified yet: a wide open space for discovery.
Methods. Mathematical modeling and algorithmic techniques imported from machine learning, AI, computational neuroscience, dynamical systems theory. Develop new mathematical languages for cognitive dynamics.
Applications. Dynamical pattern recognition (speech, text, handwriting), robot motor control, or any other temporally defined tasks. Architectures and algorithms for low-precision stochastic parallel neural microprocessors.
Teaching. Help students to make friends with mathematical abstraction in theoretical computer science and machine learning.