Menu
Research / Theme
The Gryphon Center for A.I. and Theoretical Sciences is a purpose-led organization devoted to the study of foundational topics in artificial intelligence and applications of related mathematical methods in theoretical sciences.
Our research endeavors can be broadly classified under two main themes: artificial intelligence and mathematical methods in theoretical sciences.
Our research work is devoted to various topics in the rapidly growing field of artificial intelligence, often inspired by a blend of foundational principles and real-life applications. The projects that we have completed in the past include incorporating mathematical techniques like Uniform-Manifold-Approximation-and-Projection (UMAP) in refining conventional neural network architectures and active learning methods in model training. Uncertainty quantification frameworks in machine learning is also a major theme which characterises our recent and current research focus. Our work in A.I. has mainly found applications within the field of biomedical sciences, in particular medical physics and the domain of medical imaging and radiation oncology. Yet the foundational techniques that we study carry a certain level of universal significance, and can be applied broadly.
A secondary focus of our research lies in mathematical methods that arise in the study of diverse topics in theoretical sciences. Projects that we completed in the past include : (a) using a set of coupled nonlinear differential equations in modelling biochemical reactions in cells that can potentially inform and guide our understanding of ultra-high-dose-rate radiotherapy (b) using methods of differential geometry to construct models of black hole geometries and gravitational wave solutions in Einstein’s theory of relativity, which can be further invoked to refine our understanding of cosmology and gravitational physics. The spectrum of mathematical methods that we study has found growing resonance in the field of A.I. too, such as the use of a relatively novel class of models called physics-informed-neural networks to model differential equations that arise in theoretical physics and other areas of natural sciences.