In response, we find ourselves in a new era of challenges promulgating research
with the aim to develop science and technology that emulate how the brain works
and to utilize that knowledge for creating a paradigm shift in computation towards
cognition. New technologies such as functional magnetic resonance imaging (fMRI)
and advanced image processing have also made major inroads into neuroscience,
although the theory of understanding how our brain works is at an early stage.
Nevertheless, fundamental scientific experiments to emulate and create brain-like
behavioral characteristics are underway.
In this chapter we discuss one such approach using devices, based on the atomic
switch’s synaptic-like properties, connected in a network called the “Atomic Switch
Network” (ASN). The device is inspired, not only by its synaptic function, but also
by the brain’s inherent network characteristics in the neocortex and its distributed
memory.
2 Frameworks for Neuromorphic and Bio-inspired
Computing
In order to advance the field of computing, new computational hardware paradigms
are required including artificial neural networks, deep learning, reservoir computation, and neuromorphic computing. Additionally, identifying ways to apply these
systems for use with established algorithms such as deep learning will play a key role
in ushering in the next generation of computing [7].
2.1 Neuromorphic Computing and Artificial Neural
Networks
Beautifully designed by the natural world, the human brain is capable of complex
multisensory tasks and decision-making using architecture distinct from the hyperengineered grids of von Neumann computers. Modern neurobiology describes
neurons using the formulations of the Hodgkin-Huxley model [8], emulating single
neurons as a circuit of capacitors, nonlinear resistors, and a current source. Individual neurons connect to one another through synapses, creating a connectome architecture [9] or network allowing cascading ions to transmit information. Network
connectivity determines the efficiency of interneuron communication and has been
heuristically observed to follow a small-world network topology [10, 11]. Found
throughout nature from galaxy formation to sociological trends, small-world dynamics form the core of understanding complex systems nonlinearizable by current
methods. Amazingly, cognitive behavior and memory association practiced by the
brain utilize these salient features to be able to perform both computation and
information storage within a single synapse [12]. Collectives of interacting neurons
Atomic Switch Networks for Neuroarchitectonics: Past, Present, Future
203
with the aim to develop science and technology that emulate how the brain works
and to utilize that knowledge for creating a paradigm shift in computation towards
cognition. New technologies such as functional magnetic resonance imaging (fMRI)
and advanced image processing have also made major inroads into neuroscience,
although the theory of understanding how our brain works is at an early stage.
Nevertheless, fundamental scientific experiments to emulate and create brain-like
behavioral characteristics are underway.
In this chapter we discuss one such approach using devices, based on the atomic
switch’s synaptic-like properties, connected in a network called the “Atomic Switch
Network” (ASN). The device is inspired, not only by its synaptic function, but also
by the brain’s inherent network characteristics in the neocortex and its distributed
memory.
2 Frameworks for Neuromorphic and Bio-inspired
Computing
In order to advance the field of computing, new computational hardware paradigms
are required including artificial neural networks, deep learning, reservoir computation, and neuromorphic computing. Additionally, identifying ways to apply these
systems for use with established algorithms such as deep learning will play a key role
in ushering in the next generation of computing [7].
2.1 Neuromorphic Computing and Artificial Neural
Networks
Beautifully designed by the natural world, the human brain is capable of complex
multisensory tasks and decision-making using architecture distinct from the hyperengineered grids of von Neumann computers. Modern neurobiology describes
neurons using the formulations of the Hodgkin-Huxley model [8], emulating single
neurons as a circuit of capacitors, nonlinear resistors, and a current source. Individual neurons connect to one another through synapses, creating a connectome architecture [9] or network allowing cascading ions to transmit information. Network
connectivity determines the efficiency of interneuron communication and has been
heuristically observed to follow a small-world network topology [10, 11]. Found
throughout nature from galaxy formation to sociological trends, small-world dynamics form the core of understanding complex systems nonlinearizable by current
methods. Amazingly, cognitive behavior and memory association practiced by the
brain utilize these salient features to be able to perform both computation and
information storage within a single synapse [12]. Collectives of interacting neurons
Atomic Switch Networks for Neuroarchitectonics: Past, Present, Future
203