Recent experimental findings appear to confirm that the nature of the states governing synaptic plasticity is discrete rather than continuous. This means that learning models based on discrete dynamics have more chances to provide a ground basis for modelling the underlying mechanisms associated with plasticity processes in the brain. In this paper we shall present the physical implementation of a learning model for Spiking Neural Networks (SNN) that is based on discrete learning variables. After optimizing the model to facilitate its hardware realization it is physically mapped on the POEtic tissue, a flexible hardware platform for the implementation of bio-inspired models. The implementation estimates obtained show that is possible to conceive a large-scale implementation of the model able to handle real-time visual recognition tasks.