The problem of nonlinear optimization is encountered while fitting the observed response of certain radiation detectors. In this paper the development of an optimizing technique based on quasi random search is described. The performance of this technique is compared with other existing methods of gradient search and linear Taylor differential correction technique in the case of fitting the photopeak responses of Nal(Tl) detectors. Our results are favourable especially when the observed data are associated with poor statistics.

A numerical study of nonlinear least square data fitting using random numbers from the congruential generator and several quasi-random generators is presented. The results indicate that at least up to five dimensions some of the quasi-random sequences yield better accuracy than the congruential pseudo-random sequence. Some recommendations for selecting the generators of quasi-random sequences are also given.

A study of effective implementation of threshold activation technique for neutron spectral analysis in the environment of an alpha-cyclotron target is presented. The activation data are analysed using LOUHI-82 code. Optimal choices of the regularization parameters of the code are studied and discussed. Energy distribution of neutrons emitted from thick targets of Be, C and Ta irradiated by 40–45 MeV alphas is discussed.