Page 50 - EE Times Europe September 2021

50 EE|Times EUROPE

Innatera Unveils Neuromorphic AI Chip to Accelerate Spiking Networks

APPLICATION-SPECIFIC
Innatera, a spin-out from the Delft University of Technology, was
already working with revenue customers on its SNN algorithms before
moving into hardware and raising a seed round of €5 million (about
US$6 million) toward the end of 2020.
“We’ve been working with a number of customers since the time we
actually started the company, and these engagements are still ongoing
— they’ve matured very significantly,” Kumar said. “We hope to be able
to show more demonstrations together with some of these customers
in the later part of this year.”
Kumar said the company maintains its focus as a compute solutions
company; that is, it will supply turnkey solutions that include both
hardware and application-specific SNN algorithms.
Innatera’s first chip is suitable for audio, health, and radar applica-
Innatera’s spiking neural processor includes a massively parallel tions. The company’s roadmap could include further-optimized chips
neuro-synaptic array and spike encoders and decoders. for each of the applications.
(Source: Innatera)

The second reason is to optimize performance. Rather than represent
information as bits in words, in an SNN, information is represented as
precisely timed spikes. The timing of the spikes needs to be manipu-
lated at a very fine-grained level to extract insights about the data. The
neurons and the connections between them (the synapses) therefore
need to exhibit complex timing behaviors. Those behaviors can be
adjusted via Innatera’s SDK to optimize performance.
Innatera describes its chip as analog-mixed signal or “digitally
assisted analog.” Neurons and synapses are implemented in analog
silicon to maintain ultra-low power consumption. Analog electron-
ics also allow continuous time networks (digital electronics would
require discretization). This is
important to SNNs because they
inherently have a notion of time
and must be able to hold partic-
ular states over a period of time.
“Doing this is much easier in
the analog domain — you don’t
have to shift the complexity of
keeping state into the network
topology,” Kumar said. “Our com-
pute elements naturally retain
that state information. This is
the reason we do things in the
analog domain.”
Minor inconsistencies in A compute segment in Innatera’s array, where the neurons are
Innatera’s Sumeet Kumar fabrication between compute designed to be carefully matched. Programmable synapses are
elements on the chip, and arranged in a multi-level crossbar structure. (Black lines/dashes
between different chips, can be a problem for implementing neural here represent input and output spikes.) (Source: Innatera)
networks accurately in the analog domain. Innatera’s solution
is to group neurons into what it calls segments, which are carefully
designed to match path lengths and numbers of neurons. “We architected the device in such a way so that we could
The segment design “essentially allows us to use the best of accelerate a wide variety of spiking neural networks,” Kumar said.
analog circuitry while minimizing these non-idealities that you “[Our chip] can implement these networks across application domains.
would typically have in an analog circuit,” Kumar said. “All of this But as we go deeper into domains, it may be necessary to optimize the
was essentially done to make sure that neurons inside a segment hardware design, and this is something which we will look at in the
exhibit deterministic behavior and that they function in a way that future. Right now, the hardware is not overly specialized toward any
is similar to their immediate neighbors.” specific class of applications or any style of spiking neural networks. The
Inconsistencies between chips can cause problems when the same aim is to support a variety of them generally inside the architecture.”
trained network is rolled out to devices in the field. Innatera gets Samples of the initial chip are on track to become available before
around this with software. the end of 2021. ■
“Mismatch and variability are dealt with deep inside the SDK,”
Kumar said. “If you are a power user, we can expose some of that to Sally Ward-Foxton is editor-in-chief of EE Times Weekend. This article
you, but a typical programmer doesn’t need to bother about it.” originally appeared on EE Times.

SEPTEMBER 2021 | www.eetimes.eu