An Electro-optical Connectome Prototype for Eight Neuron Representations in FPGA Technology Lorenzo Ferrara, Alexey Petrushin and Axel Blau Dept. of Neuroscience and Brain Technologies (NBT), Italian Institute of Technology (IIT), 16163 Genoa, Italy Keywords: Brain-inspired Computation, Nervous System Emulation, Optical Connectome, Parallel Information Flow, Structured Illumination, Replica-casting, Field-programmable Gate Arrays. Abstract: In nature, interneural signaling is highly parallel and temporally precisely structured. It would require equal parallelism and temporal accuracy to faithfully mimic neural communication in hardware representations. Light-based communication schemes fulfil this prerequisite. We report on a prototype of an optical connectome implementation for a neuromorphic system eventually consisting of eight neurons. The platform is based on field-programmable gate arrays (FPGAs) that run neuron-specific response models. Their axons are represented by light-emitting diodes (LEDs) with axonal arbors in the form of micro- patterned transparencies. They distribute membrane voltage threshold crossings, which are represented by light pulses, onto synapse-specific photodiodes of postsynaptic neurons. This contribution sketches out the overall system design and discusses its prospective application in replicating the connectome of the nematode C. elegans in the framework of the Si elegans project. 1 INTRODUCTION Surprisingly, even simple biological neural networks can outperform today’s fastest computational systems in tasks such as pattern recognition and locomotion control. Nervous systems are complex, highly parallel information processing architectures made of seemingly imperfect and slow, yet exceptionally adaptive and power-efficient components to carry out sophisticated information processing functions. However, despite the rapidly growing body of knowledge on almost every aspect of neural function, currently no computational model or hardware emulation exists that is able to describe or even reproduce the complete behavioural repertoire of the nematode Caenorhabditis elegans, an organism with one of the simplest known nervous systems. C. elegans, a soil-dwelling worm with a life span of a few days, 1 mm long and 80 µm in diameter, is one of the five best characterized organisms. It is multicellular and develops from a fertilized egg to an adult worm similar to higher organisms. The morphology, arrangement and connectivity of each cell including its neurons have been completely described and are found to be almost invariant across different individuals. Initially, 6393 chemical synaptic connections, 890 electrical junctions, and 1410 neuromuscular junctions were identified (White et al., 1986). Recent revisions of the original electron microscopy datasets suggest that these numbers may actually be higher. All of this data including the connectome, the detailed interconnectivity map of the 302 neurons through synapses, is publicly available through the Worm Atlas (Achacoso and Yamamoto, 1992; Oshio et al., 2003; Varshney et al., 2011). Despite its simplicity, the nervous system of C. elegans does not only sustain vital body function, but generates a rich variety of behavioural patterns in response to internal and external stimuli. These include associative and several forms of nonassociative learning that persist over several hours (Hobert, 2003). Interestingly, many processes of learning and memory in C. elegans were highly conserved across different species during evolution, which demonstrates that there are universal mechanisms underlying learning and memory throughout the animal kingdom (Lin and Rankin, 2010). To replicate the parallel information processing pathways in nervous systems as faithfully as possible, an equally parallel information Ferrara, L., Petrushin, A. and Blau, A.. An Electro-optical Connectome Prototype for Eight Neuron Representations in FPGA Technology. In Proceedings of the 3rd International Congress on Neurotechnology, Electronics and Informatics (NEUROTECHNIX 2015), pages 127-132 ISBN: 978-989-758-161-8 Copyright c  2015 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved 127