Machine learning (ML) assumes a central place in artificial intelligence (AI) and computer science, which concentrates on the utility of learning algorithms and data to simulate the human learning process. The learning performance of an ML framework has been gradually improved (e.g., by acquiring more data and using modern deep learning methods) to approach human-level performance. At VinAI, our ML group conducts cutting-edge fundamental research, which subsequently drives progress in essential applications such as computer vision, natural language processing, robotics, smart mobility, human behavior understanding, or machine translation. We question the core of intelligence, the effective learning mechanisms using data and prior knowledge, and how to translate them into efficient algorithmic implementations.
In particular, we focus on learning algorithms that can achieve (near) human-level capability in transfer learning and multi-task learning and pioneer some of the most advanced methods using optimal transport and mathematical optimization in ML. Some of our specific research areas, but not limited to, include:
- Deep generative models
- Representation learning
- Optimal transport
- Continual learning
- Robust and trustworthy ML
- Adversarial ML
- Transfer learning and Domain adaptation
