Date of Final Oral Examination (Defense)
Type of Culminating Activity
Master of Science in Electrical Engineering
Electrical and Computer Engineering
Kristy Campbell, Ph.D.
Elisa Barney Smith, Ph.D.
Kurtis Cantley, Ph.D.
The design and synthesis of artificial learning systems has been aided by the study of biological learning systems. Classic biological learning is driven by the strengthening and weakening of the synapses that connect neurons within the brain through a phenomenon known as Spike-Timing-Dependent-Plasticity. That is, synaptic connectivity between neurons is modulated by the relative timing of their spiking outputs. Similarly, neuromorphic computing architectures can implement a mesh of artificial neurons interconnected by a network of artificial synapses to mimic the learning behaviors found in nature.
Memristors, two-terminal devices whose resistance can be programmed as a function of voltage and current, offer a promising biomimetic solution for a hardware-based artificial synapse. This work focuses on characterizing the switching behavior of an ion-conducting, chalcogenide-based resistive memory in a test environment emulating the behavior of a two-neuron, single-synapse neuromorphic circuit to demonstrate learning at speeds significantly faster than those found in biological synapses.
The results from this study show that the ion-conducting memristors used in this work exhibit effective learning at time scales ranging over several orders of magnitude: from the biologically-relevant millisecond region to the faster-than-nature nanosecond region.
Drake, Kolton T., "Biomimetic Application of Ion-Conducting-Based Memristive Devices in Spike-Timing-Dependent-Plasticity" (2015). Boise State University Theses and Dissertations. 1001.
Available for download on Sunday, August 27, 2017