Identification of mathematical model of bacteria population under the antibiotic influence (Q1615419)

This paper is concerned with parameter identification for a 2-dimensional ODE model for the occurrence of resistance to antibiotics in a bacterial population. Having in mind that the process under consideration consists in multiple stages and that not all parameters can be determined experimentally, the unknown parameters are determined by solving a series of simpler inverse problems. These problems are then solved by using multiple methods, namely the gradient method, a genetic algorithm and by the Nelder-Mead method, using a mixture of real data and presumed data. A comparative study regarding the accuracy of each method employed and the necessary number of iterations is also performed, the authors concluded that the interpolation of the initial data is effective in the case of high noise if the solutions of the system are smooth enough. Also, it is determined that, for this problem, the genetic algorithm is generally inferior to the gradient method as far as the accuracy is concerned, but requires significantly fewer iterations.