The research efforts in neuromorphic engineering have yielded a wide range of impressive results, results that have seemed relevant for low-power applications interacting in a human environment. Transitioning these techniques from impressive science results to engineering solutions solving real world problems is a primary challenge in this field. We will talk about one such direction in this presentation, particularly using groups of biological neurons for classification. Recent work showing the similarity between transistors and biological channels enables a group of dense devices, spiking neuron models, programmable and adaptive synapses, and programmable dendritic trees.

These systems require analog large-scale processing and infrastructure, including enabling ultra-low power analog approaches to be programmable and configurable similar to digital solutions. Most of the techniques have only recently been developed and historically been a central focus in the neuromorphic field.   We will discuss how the analog circuit and signal processing techniques have progressed, including discussing our research at Georgia Tech on programmable analog circuits, programmable system design, and configurable systems.  We will then present possible roadmaps of analog signal processing technology going forward, as well as the upcoming challenges are for these approaches to reach the level of design complexity we currently expect for digital computation.