MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074352 A) filed by Vellore Institute Of Technology on June 16, 2026, for Neural Network System For Multi-Species Ecosystem Parameter Estimation And Stability Analysis.
Inventors include Dr T Poornima; and G Ramraj.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT NEURAL NETWORK SYSTEM FOR MULTI-SPECIES ECOSYSTEM PARAMETER ESTIMATION AND STABILITY ANALYSIS A computer- implemented method (100) for ecosystem parameter estimation and stability analysis comprises receiving (102) sparse time-series wildlife population observations for prey, juvenile predator, adult predator, and scavenger species. The method comprises formulating (104) a four-species reaction-diffusion partial differential equation system incorporating a fear factor, ontogenetic stage structure distinguishing juvenile and adult predators, a maturation time delay, and scavenger dynamics. The method comprises building (106) a physics-informed neural network having hidden layers with Swish activation functions, residual connections, and a softplus output mapping time to predicted species abundances. The method comprises assembling (108) a composite loss combining data, partial differential equation residual, and initial-condition loss terms, training (110) the neural network using two-stage optimisation that jointly fits network weights and kinetic parameters, and quantifying (112) uncertainty using Monte-Carlo dropout to produce confidence intervals.
Disclaimer: Curated by HT Syndication.
