xiv
Preface
This is in turn particularly relevant to future computing needs such as cognitive
processing, big-data analysis, and low-power intelligent systems based on the
Internet of Things.
Neuromorphic/synaptic electronics is an emerging field of research aiming to
overcome Von Neumann platforms by building artificial neuronal systems that
mimic the extremely energy-efficient biological synapses. Neuromorphic memristive architectures integrated into edge computing devices are expected to increase
the data processing capability at lower power requirements and reduce several overheads for cloud computing solutions [10]. The introduction of photovoltaic/photonic
aspects into neuromorphic architectures could produce self-powered adaptive electronics and open new possibilities in artificial neuroscience, neural communications,
sensing, and machine learning. This would enable, in turn, a new era for computational systems owing to the possibility of attaining high bandwidths with much
reduced power consumption [11].
Memristor devices operated as a nonvolatile memory (NVM) are emerging for
data storage and unconventional computing systems. In this case, a memristor
should display two (or more) largely different values of memristance and be a
nonvolatile device. The memristor is driven from one memristance to others via
a suitable current (or voltage) pulse, an operation that is often referred to as
set/reset. This mechanism permits to exploit nonvolatile resistive states of memristor
to encode information bits. A continuous range of resistive states enables the
use of memristor devices as analogue programmable resistors. Such devices,
whose resistance can be precisely modulated electronically, and which can support
important synaptic functions (e.g., Spike-Timing-Dependent Plasticity—STDP),
pave the way to denser low-power analogue circuits, multi-state memory, and
large-scale synapse implementation in neuromorphic systems and cellular neural
networks with adaptation capabilities [10, 12].
Memristor devices embedded in nonlinear circuits can generate a complex
nonlinear dynamical evolution of their memristance that can be exploited to build
nanoscale oscillators potentially described by low-order mathematical/circuit models. Networks of interconnected and interacting oscillators can develop cooperative
and collective dynamics, e.g., phase synchronization and other self-organizing spatiotemporal phenomena for alternative computing schemes overpassing the limits
of conventional digital and Boolean computation. A memristor-based nonlinear
oscillator is proposed in [13] as a source of tunable chaotic behavior that can
be incorporated into a Hopfield computing network to improve the efficiency
and accuracy of converging to a solution for computationally hard problems. The
breakthrough is the development of networks of interacting nanoscale memristor
oscillators and their computational schemes based on cooperative and collective
dynamics. It is then crucial to develop a technological platform comprehensive of
functional materials and hardware memristors including physical/circuit models of
their operation as a key in hand tool for the large-scale industrial exploitation.
Preface
This is in turn particularly relevant to future computing needs such as cognitive
processing, big-data analysis, and low-power intelligent systems based on the
Internet of Things.
Neuromorphic/synaptic electronics is an emerging field of research aiming to
overcome Von Neumann platforms by building artificial neuronal systems that
mimic the extremely energy-efficient biological synapses. Neuromorphic memristive architectures integrated into edge computing devices are expected to increase
the data processing capability at lower power requirements and reduce several overheads for cloud computing solutions [10]. The introduction of photovoltaic/photonic
aspects into neuromorphic architectures could produce self-powered adaptive electronics and open new possibilities in artificial neuroscience, neural communications,
sensing, and machine learning. This would enable, in turn, a new era for computational systems owing to the possibility of attaining high bandwidths with much
reduced power consumption [11].
Memristor devices operated as a nonvolatile memory (NVM) are emerging for
data storage and unconventional computing systems. In this case, a memristor
should display two (or more) largely different values of memristance and be a
nonvolatile device. The memristor is driven from one memristance to others via
a suitable current (or voltage) pulse, an operation that is often referred to as
set/reset. This mechanism permits to exploit nonvolatile resistive states of memristor
to encode information bits. A continuous range of resistive states enables the
use of memristor devices as analogue programmable resistors. Such devices,
whose resistance can be precisely modulated electronically, and which can support
important synaptic functions (e.g., Spike-Timing-Dependent Plasticity—STDP),
pave the way to denser low-power analogue circuits, multi-state memory, and
large-scale synapse implementation in neuromorphic systems and cellular neural
networks with adaptation capabilities [10, 12].
Memristor devices embedded in nonlinear circuits can generate a complex
nonlinear dynamical evolution of their memristance that can be exploited to build
nanoscale oscillators potentially described by low-order mathematical/circuit models. Networks of interconnected and interacting oscillators can develop cooperative
and collective dynamics, e.g., phase synchronization and other self-organizing spatiotemporal phenomena for alternative computing schemes overpassing the limits
of conventional digital and Boolean computation. A memristor-based nonlinear
oscillator is proposed in [13] as a source of tunable chaotic behavior that can
be incorporated into a Hopfield computing network to improve the efficiency
and accuracy of converging to a solution for computationally hard problems. The
breakthrough is the development of networks of interacting nanoscale memristor
oscillators and their computational schemes based on cooperative and collective
dynamics. It is then crucial to develop a technological platform comprehensive of
functional materials and hardware memristors including physical/circuit models of
their operation as a key in hand tool for the large-scale industrial exploitation.