![](bg2.png)
Figure 1: Connectivity matrix of currently 275 neurons
extracted from the Neural Connectivity II dataset by
Varshney et al. (Varshney et al., 2011). Pre-synaptic
neurons listed in the row headers connect to their post-
synaptic target neurons listed in the column headers via
one or up to three simultaneous synaptic connections. Due
to page size limitations, the names of the individual
neurons are not legible. Colour codes for the number of
synaptic connections: 1: yellow, 2: green, 3 red. Electrical
gap junctions and neuromuscular junctions are not
included in this matrix.
3 THE Si elegans CONNECTOME
One of the key challenges and features of the Si
elegans computational platform is the development
and implementation of the 3D electro-optical
interconnectivity concept for the parallel processing
and transmission of neuronal information. In current
2D interconnectivity designs based on static
integrated circuitry-only schemes, network
complexity is limited by 2D interconnectivity
bottlenecks. Thus, inter-neuron connectivity is a
major problem. To circumvent this limitation, other
groups have deployed shared bus-based connectivity
concepts and asynchronous address-based event-
coding/event-representation systems (AES, AER,
NoC, ..) to mimic parallel information transmission.
But they are serial in nature. As long as processing
rates are sufficiently high for a low number of
synaptic connections (several thousands), their serial
nature can be hidden and parallelism be pretended.
However, a system may encounter communication
bottlenecks when a high number of target synapses
need to be addressed simultaneously. In that case,
non-parallelism will become apparent. A more
serious problem of serial-type simulations is their
stochastic jitter in the timing of events that prevent
the accurate and reproducible mimicry of parallel
information flow between neurons.
Figure 2: Concept and elements of an individual Si
elegans FPGA neuron module and its comparison with a
real neuron. Neural activity will arrive at individual input
lines of an FPGA (i) and will be processed by the neuron-
specific stimulus-response algorithm that the FPGA was
programmed with (ii). Its output activity will be
distributed in parallel through signal distribution elements
(iii) to individual input lines (i) of the target FPGA
neurons to which the neuron connects to. In case of signal
propagation by light, incoming activity will arrive as
spatially confined light pulses at individual pixels of
photoelectric converters (synapses/gap junctions) being
individually connected to the individual FPGA input lines.
Neural response activity generated by the neural model
residing on the FPGA will trigger a coherent light source
at one of its output lines (axon). This light will pass
through light distribution elements to distribute activity
onto selected pixels (synapses/gap junctions) of inter-
connected target neurons. In case of electrical signal
transmission through wires, a split-wire bundle will
transmit a digital signal pulse from the axonal output line
of the FPGA to individual synapse/gap junction sockets of
the target neurons. Reproduced with permission from the
Si elegans project consortium.
These limitations will be overcome through
research into free-space communication techno-
logies. This concept picks up on research into free-
space optical computation and build on rapid
progress in optical communication technology, be it
for telecommunication or, more recently, on-chip
and intra-chip optical interconnects (Assefa et al.,
2010; Doany et al., 2012; Loughran, 2010; Orcutt et
al., 2011). The C. elegans connectome will be
replicated by connecting the individual FPGA
neuron modules in a line-of-sight framework
through light. This approach allows for the
distribution of the signal at the axonal output or gap-
junction pin of an FPGA neuron onto the respective
synaptic input or gap-junction pins of those target
TowardsanElectro-opticalEmulationoftheC.elegansConnectome
185