Output details

Modeling fast stimulus-response association learning along the occipito-parieto-frontal pathway following rule instructions.

<24> The paper proposes an original learning architecture, using feedback signals guiding a feedforward push, and an original synaptic learning rule, based on the allocation of limited weight resources. This model allows for the first time interference-free one-shot learning in multilayer networks of spiking neurons. As the model implements a key mechanism of fast stimulus-response learning, it significantly widens the application domain of neuromorphic hardware. The proposed learning rule eliminates the need for artificial boundaries to weights and can be incorporated in a number of existing learning models.