Organic neuromorphic devices: Past, present, and future challenges
- PDF / 2,192,036 Bytes
- 12 Pages / 585 x 783 pts Page_size
- 61 Downloads / 212 Views
Neuromorphic computing With the ultimate aspiration of crafting an “artificial brain,” since the earliest stages of development, the fields of computer science and neurobiology have been intricately linked, as advances in each field have motivated reconsideration in the other.1,2 With the advancement of very large-scale integration (VLSI) frameworks, neuromorphic systems were introduced in the late 1980s to early 1990s as a methodology by which computers could better emulate the brain’s natural processes, and thereby harness the associated advantages in power consumption and complexity of computation.3 In the intervening decades, much of the work toward a synthetic neural system has been focused in one of two directions. For the purpose of distinction, we will use the terms simulation and emulation to refer to two distinct processes. Simulation is the process by which phenomenon A is replicated
phenomenologically with no regard to how the implementation may or may not mimic the original process. A familiar example comes from the world of Monte Carlo simulations, wherein the underlying physics need not be hard-coded into the simulation in order to accomplish something useful. On the other hand, emulation is the process by which phenomenon A is replicated with an eye toward reproducing the processes of A in the most faithful way possible. Consider, for example, the hardware limitations that are put in place by console designers in order to ensure backwards compatibility, thereby emulating an earlier generation of devices. Along similar lines, we can consider multiple avenues by which neural processes have been pursued in the past decades. Inspired by the feed-forward nature of neural signals, artificial neural networks were initially designed to simulate neural signal processing by way of a high-level analogy between
Yaakov Tuchman, Department of Materials Science, Stanford University, USA; [email protected] Tanyaradzwa N. Mangoma, Department of Engineering, University of Cambridge, UK; [email protected] Paschalis Gkoupidenis, Department of Molecular Electronics, Max Planck Institute for Polymer Research, Germany; [email protected] Yoeri van de Burgt, Neuromorphic Engineering Group, Eindhoven University of Technology, The Netherlands, [email protected] Rohit Abraham John, School of Materials Science and Engineering, Nanyang Technological University, Singapore; [email protected] Nripan Mathews, School of Materials Science and Engineering, Nanyang Technological University, Singapore; [email protected] Sean E. Shaheen, Department of Electrical, Computer, and Energy Engineering, and Department of Physics, University of Colorado Boulder, USA; [email protected] Ronan Daly, Institute for Manufacturing, Department of Engineering, University of Cambridge, UK; [email protected] George G. Malliaras, University of Cambridge, UK; [email protected] Alberto Salleo, Stanford University, USA; [email protected] *denotes Guest Editor of this issue doi:10.1557/mrs.2020.196 • VOLUME • AUGUST © 2020 Mater
Data Loading...
