580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can user-friendly interfaces and algorithmic enhancements improve the efficiency and accessibility of ODE boundary value problem solvers?
This research theme focuses on the development of boundary value problem (BVP) solvers for ordinary differential equations (ODEs) that enhance both the user experience through simplified interfaces and the solver's performance via algorithmic advancements. The significance lies in lowering the barrier for users to apply sophisticated BVP solvers while handling complex problems such as singular coefficients and unknown parameters effectively.
Key finding: This paper introduced BVP_SOLVER, a modern Fortran 90/95 code evolved from the widely-used MIRKDC package, which significantly simplifies the user interface by reducing the number of arguments and subroutines the user must... Read more
2. What numerical solver strategies best achieve a balance between accuracy, efficiency, and stability for initial value problems (IVPs) in ODEs with variable stiffness and discontinuities?
This theme investigates a variety of numerical algorithms for initial value ordinary differential equations, particularly under challenges such as variable stiffness and state discontinuities (e.g., pulses or threshold crossings). Efficient solvers must adapt their step sizes appropriately, accurately detect transient features within the solution, and maintain computational speed, especially on limited arithmetic hardware like fixed-point processors. Exploring implicit vs. explicit methods and solver parameter tuning is vital for achieving these goals.
Key finding: The paper develops software that uses defect sampling to detect short-duration, abrupt pulses in ODE IVPs without modifying existing solvers. This method evaluates continuous solution approximations within solver steps and... Read more
Key finding: This work demonstrates that fixed-point neural ODE solvers (as used in SpiNNaker hardware) require balancing speed and accuracy by employing techniques like explicit solver reduction and correction mechanisms for accumulating... Read more
Key finding: Through evaluation of MATLAB’s seven built-in ODE solvers on a heart valve model exhibiting changing stiffness, this study elucidates the impact of solver type, relative tolerance, and initial step size on computational cost... Read more
Key finding: The authors develop a nonlinear interpolating function combining polynomial, exponential, and trigonometric components to formulate a novel numerical integrator for first-order IVPs. The method satisfies Lipschitz conditions... Read more
3. How can advanced algorithmic frameworks and equivalence relations enable model reduction and efficient solving of large ODE systems?
This theme addresses the systematic reduction of large ODE systems by identifying variable equivalences and applying algorithmic partition refinement methods. By collapsing variables with identical or aggregate behaviors, these approaches yield smaller, computationally tractable systems that preserve essential dynamics. Tools implementing such theories are crucial for practical analysis and simulation of complex biochemical and dynamical networks.
Key finding: ERODE implements forward and backward differential equivalences that aggregate or identify identical ODE variables, respectively, reducing system dimension without losing accuracy. For polynomial systems up to degree two, the... Read more