Neural Network Topology Formation Using Memristive Jaumann Structures

The interconnection of neurons through synapses are essential in the context of learning behavior. This includes the topology formation process in neural networks, i.e., the synaptic coupling of neurons. Memristors are important devices in this context, especially since the formation process must be dynamic. Attaching memristors to a Jaumann structure allows for dynamic topologies that can contain unidirectional couplings. These memristors can be specifically and independently manipulated with this structure as well. Although this work investigates supervised topology formation, the proposed element is a potential candidate for self-organizing systems.