The presence of nonlinearity in the DNLS equation leads to reduction of energy or charge transfer from one site to others. During the chemical processes in photosynthesis, we have an excitation interacting with vibrational modes that tunnels in an ultrafast way. The targeted energy transfer model is based in the DNLS equation and provides a mechanism for fast transfer between sites. The basic feature of the model is that it includes disorder in both the local energy sites and the nonlinearity parameters that are constrained in a specific way. This leads to a nonlinear resonance that enhances transfer and can be analyzed analytically. We use machine learning methodology in order to discover this resonance both in the solvable dimer case as well as the trimer case where there is no analytical solution available. Finding nonlinear resonances through machine learning is quite useful in the study of complex systems.