In machine learning and data science, we have experienced a surge in the amount of available data in recent years. Right now, we witness a shift towards data-intensive and evidence-based decision making with a radical impact across many aspects of science and commerce. In order to adapt to this new realm, commercial applications are in desperate need of intelligent systems. Yet, we are only at the beginning of this evolution that increasingly imposes new demands such as computationally tractable algorithms, algorithms that protect privacy issues, and learning methods that are capable of exploiting huge amounts of unlabeled data.
Our research focuses on pattern recognition, machine learning, and computational data analytics with applications in various fields, ranging from signal and speech processing to medical data analysis and data modeling problems in industrial applications. Our vision is to bridge the gap between basic research, applications, and intelligent systems, as shown in the Figure. This has a mutual benefit across all aspects of machine learning methods as real-world problems inspire the development of novel methods and vice versa. We are particularly interested in probabilistic graphical models for reasoning under uncertainty, discriminative and hybrid learning paradigms, deep learning, and sequence modeling.
Despite its practical and commercial success, many aspects remain under-explored and provide rich research opportunities. Ultimately, the goal is to construct systems that continually improve through experience. We aim to contribute to the knowledge of modeling, learning, and reasoning of complex large-scale data. This research will enable and foster services for science and society including health care, manufacturing, and education.
Currently we are active in the following fields:
- Hardware-aware Machine Learning
- Computational Medicine
- Probabilistic Graphical Models and Belief Propagation
- Speech Separation and Dereverberation
- Predictive Maintenance (several Industry Projects)
- Radar Signal Denoising