Training physical properties of quantum systems

The strategy of training (an efficient search in a huge parameter space) quantum systems to fulfill some nontrivial physical properties is addressed, as opposed to the usual approach of examining physical properties of a given Hamiltonian. The training process is exemplified by two- and three-dimensional Anderson models. Results indicate that small changes in the parameter space affect dramatically the statistical properties of the level spacing of three-dimensional systems in the insulating regime. Other applications of the training strategy are also briefly discussed.