A fractional-order model for MINMOD Millennium

Identifiability and predictability of integer- and fractional-order epidemiological models using physics-informed neural networks

We analyze a plurality of epidemiological models through the lens of physics-informed neural networks (PINNs) that enable us to identify time-dependent parameters and data-driven fractional differential operators. In particular, we consider several variations of the classical susceptible-infectious-removed (SIR) model by introducing more compartments and fractional-order and time-delay models. We report the results for the spread of COVID-19 in New York City, Rhode Island and Michigan states and Italy, by simultaneously inferring the unknown parameters and the unobserved dynamics. For integer-order and time-delay models, we fit the available data by identifying time-dependent parameters, which are represented by neural networks. In contrast, for fractional differential models, we fit the data by determining different time-dependent derivative orders for each compartment, which we represent by neural networks. We investigate the structural and practical identifiability of these unknown functions for different datasets, and quantify the uncertainty associated with neural networks and with control measures in forecasting the pandemic.

Modeling continuous glucose monitoring with fractional differential equations subject to shocks

Continuous Glucose Monitoring (CGM) produces long time-series of noisy observations of a single variable (tissue glucose concentration), whose evolution may be explained by a dynamical model. In order to represent the unknown mixture of possible control mechanisms of different orders affecting the measured variable, a fractional differential approach seems justified. In any case, variations in food intake and/or physical activity ought to be taken into account if a plausible interpretation of the dynamics is to be obtained. In the present work, the mathematical construction and the numerical implementation of a Fractional Differential Equations (FDE) initial value problem are systematically reviewed, with the intent of offering the reader a concise and mathematically rigorous description of this approach. An FDE model for CGM is formulated: the model includes compartments for stomach and intestinal glucose contents and for blood and tissue (subcutaneous) glucose concentrations, as well as the shock effects of food ingestion and of increased glucose consumption due to physical activity. The model parameters, including the (non-integer) order of differentiation, are estimated from CGM observations on six Type 1 diabetic patients. The best-fit fractional orders for the six subjects range from 1.59 to 2.13. For comparison, best fits have also been computed for all subjects using an average fractional order of 1.9 and integer orders of 1 and 2.The results indicate that in the case of CGM the fractional differential model, which should be physiologically more appropriate, in fact fits the data much better than the first-order model and also better than the 2nd-order model.