In this study, we use the shape memory material to construct a tunable vibration absorber and to measure its dynamic frequency response characteristics in different temperature setting. Attaching the tunable vibration absorber to the main system of a cantilever beam, we apply adaptive sliding mode control to change its stiffness and damping ratio by heating/cooling the vibration absorber. In addition, a neural network is trained to model the inverse dynamic behavior of the tunable vibration absorber. Therefore, an adaptive and robust vibration suppression method can be obtained by using adaptive sliding mode control with a neural network to tune the natural frequency of the absorber to avoid system resonance.