AI systems, unlike humans, are brittle, not robust, often struggle when faced with novel situations, and highly sensitive to small perturbations, which can lead to catastrophically poor performance.
My research aims to develop trusted and safe machine intelligence, by building tools to address learning framework, algorithmic, data, and computing challenges.
My perspective is to connect computing issues in representation
learning (imperfect data, structure knowledge), self-supervised learning (limitation of labels), interactive learning (weak supervision and uncertain environments), adaptive learning (shift and drifted environment), stream learning (limitation of memory) and more as disruption-robust learning.
This connection not only provides a unified understanding, but also paves a principled and innovative way to design trusted and safe systems as a disruption-robust framework.
I execute two important steps steps towards this vision.
The first step (data representation construct) aims to integrate structure knowledge, self-optimization, explainability to achieve deep robust representation to fight imperfect and complexity data.
The second step (learning strategy construct) aims to integrate robust representations with adaptive and interactive learning to fight uncertain and constrained environments.