RRAM-Based Neuromorphic Computing
Systems
Putu Andhita Dananjaya, Roshan Gopalakrishnan, and Wen Siang Lew
Abstract New computational paradigms have been widely investigated in order
to further improve the approach in handling the exponentially increasing amount
of data generated across the globe as well as various emerging hardware requirements to execute complex tasks, e.g., pattern recognition, speech classification, etc.
Neuromorphic computing has emerged as one of the most extensively investigated
among these approaches. RRAM devices with their desired characteristics have been
rigorously engineered to meet the synaptic element requirements to realize such
robust yet power efficient platform. Among the performance parameters necessary
to achieve an ideal synaptic device in context of RRAM device, there are certain
dependency and potential trade-offs. In this chapter, different type of RRAM, i.e.,
anion and cation, devices based on their underlying physical mechanism with various
advantages and disadvantages are discussed. Different techniques that have been
implemented to improve the device synaptic characteristics from material viewpoint
and programming approach followed by several system level simulations demonstrating the projected performance of these devices are provided in detail. Different
algorithms available for the RRAM synapse implementation are also discussed.
1 Introduction
Human brain performs massively parallel and low power operations. It can outperform present age microprocessors on many tasks involving pattern recognition and
input classification. The underlying neurons are heavily inter-connected; on average
each neuron is connected to 10,000 (or up to 100,000) other neurons [1–3]. Despite
P. A. Dananjaya · W. S. Lew (B)
Division of Physics and Applied Physics, School of Physical and Mathematical Sciences,
Nanyang Technological University, 21 Nanyang Link, Jurong West 637371, Singapore
e-mail: wensiang@ntu.edu.sg
R. Gopalakrishnan
Institute for Infocomm Research (I2R), Agency for Science, Technology and Research
(A*STAR), Fusionopolis Way 138632, Singapore
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021
W. S. Lew et al. (eds.), Emerging Non-volatile Memory Technologies,
https://doi.org/10.1007/978-981-15-6912-8_12
383
Systems
Putu Andhita Dananjaya, Roshan Gopalakrishnan, and Wen Siang Lew
Abstract New computational paradigms have been widely investigated in order
to further improve the approach in handling the exponentially increasing amount
of data generated across the globe as well as various emerging hardware requirements to execute complex tasks, e.g., pattern recognition, speech classification, etc.
Neuromorphic computing has emerged as one of the most extensively investigated
among these approaches. RRAM devices with their desired characteristics have been
rigorously engineered to meet the synaptic element requirements to realize such
robust yet power efficient platform. Among the performance parameters necessary
to achieve an ideal synaptic device in context of RRAM device, there are certain
dependency and potential trade-offs. In this chapter, different type of RRAM, i.e.,
anion and cation, devices based on their underlying physical mechanism with various
advantages and disadvantages are discussed. Different techniques that have been
implemented to improve the device synaptic characteristics from material viewpoint
and programming approach followed by several system level simulations demonstrating the projected performance of these devices are provided in detail. Different
algorithms available for the RRAM synapse implementation are also discussed.
1 Introduction
Human brain performs massively parallel and low power operations. It can outperform present age microprocessors on many tasks involving pattern recognition and
input classification. The underlying neurons are heavily inter-connected; on average
each neuron is connected to 10,000 (or up to 100,000) other neurons [1–3]. Despite
P. A. Dananjaya · W. S. Lew (B)
Division of Physics and Applied Physics, School of Physical and Mathematical Sciences,
Nanyang Technological University, 21 Nanyang Link, Jurong West 637371, Singapore
e-mail: wensiang@ntu.edu.sg
R. Gopalakrishnan
Institute for Infocomm Research (I2R), Agency for Science, Technology and Research
(A*STAR), Fusionopolis Way 138632, Singapore
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021
W. S. Lew et al. (eds.), Emerging Non-volatile Memory Technologies,
https://doi.org/10.1007/978-981-15-6912-8_12
383