Research Areas
Creating a Continuum
between Connectionism and Symbolism
Although symbolic reasoning can be approximated by well-designed connectionist networks, it is a common view in AI that connectionist approaches and symbolic approaches are incompatible. We target at geometrically creating a continuum between symbolic models and connectionist models. To this end, vector embeddings and back-propagation methods have to be abandoned. Symbols are embodied by balls in high dimension space. Geometric construction processes are like everyday-life route descriptions that instruct where and how symbolic balls shall be located from their initial locations. The existence of such ball configuration resolved almost all the related neural-symbol debates in the literature, and paves the way to synergistic unification of connectionism and symbolism. Main Contact: Tiansi Dong
Visual Analytics
Visual analytics supporting all stages of data science process, including validation of data quality and suitability for tasks, analysis, communication of results, and decision support. Main Contact: Gennady Andrienko
Theoretical foundations of visual analytics and visual data science; Development of novel visual analytics methods for analysis of spatial and temporal data. Main Contact: Natalia Andrienko
