The investigators hypothesize that a myoelectric (EMG) controlled VR interface allows for effective upper limb motor recovery of stroke patients. EMG control offers the possibility to alter visual feedback according to the recorded muscle activity in real-time. By manipulating the motion of a virtual hand associated with the recorded muscle patterns, assistance can be provided to stroke patients by correcting the error between the actual (dysfunctional) and a reference (functional) muscle pattern. Thus, through such an assistive EMG control algorithm, patients will be able to perform reaching movements with the virtual hand despite their motor impairment. By gradually reducing assistance, it is hypothesized that the salient error in the task space provided as visual feedback will systematically change the muscle patterns, thereby driving adaptation of the dysfunctional muscle patterns, enhancing motor recovery. Moreover, due to its relevant role in motor learning, it is expected that cerebellar stimulation will favor the underlying processes of adapting cerebello-cortical plasticity involved in motor learning. Therefore, it is hypothesized that an assistive EMG control algorithm in combination with cerebellar transcranial magnetic stimulation will further enhance upper limb recovery.