Neuromorphic computing: Difference between revisions

Revision as of 20:51, 22 May 2025

Latest revision as of 03:41, 8 August 2025

(5 intermediate revisions by 5 users not shown)

Line 2:

{{Use American English|date = January 2019}}

{{Use mdy dates|date = January 2019}}

{{Machine learning|Paradigms}}

'''Neuromorphic computing''' is an approach to computing that is inspired by the structure and function of the human brain.<ref>{{Cite journal |last1=Ham |first1=Donhee |last2=Park |first2=Hongkun |last3=Hwang |first3=Sungwoo |last4=Kim |first4=Kinam |title=Neuromorphic electronics based on copying and pasting the brain |url=https://www.nature.com/articles/s41928-021-00646-1 |journal=Nature Electronics |year=2021 |language=en |volume=4 |issue=9 |pages=635–644 |doi=10.1038/s41928-021-00646-1 |s2cid=240580331 |issn=2520-1131|url-access=subscription }}</ref><ref>{{Cite journal |last1=van de Burgt |first1=Yoeri |last2=Lubberman |first2=Ewout |last3=Fuller |first3=Elliot J. |last4=Keene |first4=Scott T. |last5=Faria |first5=Grégorio C. |last6=Agarwal |first6=Sapan |last7=Marinella |first7=Matthew J. |last8=Alec Talin |first8=A. |last9=Salleo |first9=Alberto |date=April 2017 |title=A non-volatile organic electrochemical device as a low-voltage artificial synapse for neuromorphic computing |url=https://www.nature.com/articles/nmat4856 |journal=Nature Materials |language=en |volume=16 |issue=4 |pages=414–418 |doi=10.1038/nmat4856 |pmid=28218920 |bibcode=2017NatMa..16..414V |issn=1476-4660}}</ref> A neuromorphic computer/chip is any device that uses physical [[artificial neuron]]s to do computations.<ref>{{cite journal|last1=Mead|first1=Carver|title=Neuromorphic electronic systems|journal=Proceedings of the IEEE|date=1990|volume=78|issue=10|pages=1629–1636|doi=10.1109/5.58356|s2cid=1169506 |url=https://authors.library.caltech.edu/53090/1/00058356.pdf}}</ref><ref name=":2" /> In recent times, the term ''neuromorphic'' has been used to describe [[Analogue electronics|analog]], [[Digital electronics|digital]], [[Mixed-signal integrated circuit|mixed-mode analog/digital VLSI]], and software systems that implement models of [[neural system]]s (for [[perception]], [[motor control]], or [[multisensory integration]]). Recent advances have even discovered ways to mimicdetect thesound humanat nervousdifferent systemwavelengths through liquid solutions of chemical systems.<ref>{{Cite journal |last1=Tomassoli |first1=Laura |last2=Silva-Dias |first2=Leonardo |last3=Dolnik |first3=Milos |last4=Epstein |first4=Irving R. |last5=Germani |first5=Raimondo |last6=Gentili |first6=Pier Luigi |date=2024-02-08 |title=Neuromorphic Engineering in Wetware: Discriminating Acoustic Frequencies through Their Effects on Chemical Waves |url=https://pubs.acs.org/doi/10.1021/acs.jpcb.3c08429 |journal=The Journal of Physical Chemistry B |language=en |volume=128 |issue=5 |pages=1241–1255 |doi=10.1021/acs.jpcb.3c08429 |pmid=38285636 |issn=1520-6106|url-access=subscription }}</ref> An article published by AI researchers at [[Los Alamos National Laboratory]] states that, "neuromorphic computing, the [[next generation]] of [[Artificial intelligence|AI]], will be smaller, faster, and more efficient than the [[human brain]]."<ref>{{Cite web |last=Dickman |first=Kyle |title=Neuromorphic computing: the future of AI {{!}} LANL |url=https://www.lanl.gov/media/publications/1663/1269-neuromorphic-computing |access-date=2025-04-16 |website=Kyle Dickman |language=en}}</ref>

A key aspect of neuromorphic engineering is understanding how the [[Morphology (biology)|morphology]] of individual neurons, circuits, applications, and overall architectures creates desirable computations, affects how [[information]] is represented, influences robustness to damage, incorporates learning and development, adapts to local change (plasticity), and facilitates evolutionary change.

Line 17 ⟶ 18:

==Implementation==

The implementation of neuromorphic computing on the hardware level can be realized by oxide-based [[memristor]]s,<ref name="Maan 1–13">{{Cite journal|last1=Maan|first1=A. K.|last2=Jayadevi|first2=D. A.|last3=James|first3=A. P.|date=2016-01-01|title=A Survey of Memristive Threshold Logic Circuits|journal=IEEE Transactions on Neural Networks and Learning Systems|volume=PP|issue=99|pages=1734–1746|doi=10.1109/TNNLS.2016.2547842|pmid=27164608|issn=2162-237X|arxiv=1604.07121|bibcode=2016arXiv160407121M|s2cid=1798273}}</ref> [[Spintronics|spintronic]] memories, threshold switches, [[transistor]]s,<ref>{{Cite journal|title = Mott Memory and Neuromorphic Devices|journal = Proceedings of the IEEE|date = 2015-08-01|issn = 0018-9219|pages = 1289–1310|volume = 103|issue = 8|doi = 10.1109/JPROC.2015.2431914|first1 = You|last1 = Zhou|first2 = S.|last2 = Ramanathan|s2cid = 11347598|url=https://zenodo.org/record/895565}}</ref><ref name=":2">{{Cite conference|authorauthor1=Rami A. Alzahrani|author2=Alice C. Parker|title=Neuromorphic Circuits With Neural Modulation Enhancing the Information Content of Neural Signaling |conference=International Conference on Neuromorphic Systems 2020|date=July 2020|pages=1–8|language=EN|doi=10.1145/3407197.3407204|s2cid=220794387|doi-access=free}}</ref> among others. The implementation details overlap with the concepts of [[Artificialartificial immune system|Artificial Immune Systems.]]s. Training software-based neuromorphic systems of [[spiking neural networks]] can be achieved using error backpropagation, e.g. using [[Python (programming language)|Python]]-based frameworks such as snnTorch,<ref>{{cite arXiv|last1=Eshraghian|first1=Jason K.|last2=Ward|first2=Max|last3=Neftci |first3=Emre|last4=Wang|first4=Xinxin|last5=Lenz|first5=Gregor|last6=Dwivedi|first6=Girish|last7=Bennamoun|first7=Mohammed|last8=Jeong|first8=Doo Seok|last9=Lu|first9=Wei D.|title=Training Spiking Neural Networks Using Lessons from Deep Learning |date=1 October 2021 |class=cs.NE |eprint=2109.12894 }}</ref> or using canonical learning rules from the biological learning literature, e.g. using BindsNet.<ref>{{Cite web |url=https://github.com/Hananel-Hazan/bindsnet | title=Hananel-Hazan/bindsnet: Simulation of spiking neural networks (SNNs) using PyTorch.| website=[[GitHub]]| date=31 March 2020}}</ref>

==Examples==

Line 27 ⟶ 28:

In June 2012, [[spintronic]] researchers at [[Purdue University]] presented a paper on the design of a neuromorphic chip using [[Spin valve|lateral spin valve]]s and [[memristor]]s. They argue that the architecture works similarly to neurons and can therefore be used to test methods of reproducing the brain's processing. In addition, these chips are significantly more energy-efficient than conventional ones.<ref name="Spin Devices Prop">{{Cite arXiv|title=Proposal For Neuromorphic Hardware Using Spin Devices|eprint=1206.3227|last1=Sharad|first1=Mrigank|last2=Augustine|first2=Charles|last3=Panagopoulos|first3=Georgios|last4=Roy|first4=Kaushik|class=cond-mat.dis-nn|year=2012}}</ref>

Research at [[HP Labs]] on Mott memristors has shown that while they can be non-[[Volatile memory|volatile]], the volatile behavior exhibited at temperatures significantly below the [[phase transition]] temperature can be exploited to fabricate a [[neuristor]],<ref name=":0" /> a biologically- inspired device that mimics behavior found in neurons.<ref name=":0">{{Cite journal | doi = 10.1038/nmat3510| pmid = 23241533| title = A scalable neuristor built with Mott memristors| journal = Nature Materials| volume = 12| issue = 2| pages = 114–7| year = 2012| last1 = Pickett | first1 = M. D. | last2 = Medeiros-Ribeiro | first2 = G. | last3 = Williams | first3 = R. S. | bibcode = 2013NatMa..12..114P| s2cid = 16271627}}</ref> In September 2013, they presented models and simulations that show how the spiking behavior of these neuristors can be used to form the components required for a [[Turing machine]].<ref>{{cite journal|doi=10.1088/0957-4484/24/38/384002|title=Phase transitions enable computational universality in neuristor-based cellular automata|author1=Matthew D Pickett|author2=R Stanley Williams|name-list-style=amp|date=September 2013|publisher=IOP Publishing Ltd|journal=Nanotechnology|volume=24|issue=38|pmid=23999059|bibcode=2013Nanot..24L4002P|s2cid=9910142 |at=384002}}</ref>

[[Neurogrid]], built by ''Brains in Silicon'' at [[Stanford University]],<ref>{{cite journal|last1=Boahen|first1=Kwabena|title=Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations|journal=Proceedings of the IEEE|date=24 April 2014|volume=102|issue=5|pages=699–716|doi=10.1109/JPROC.2014.2313565|s2cid=17176371}}</ref> is an example of hardware designed using neuromorphic engineering principles. The circuit board is composed of 16 custom-designed chips, referred to as NeuroCores. Each NeuroCore's analog circuitry is designed to emulate neural elements for 65536 neurons, maximizing energy efficiency. The emulated neurons are connected using digital circuitry designed to maximize spiking throughput.<ref>{{cite journal|doi=10.1038/503022a|pmid = 24201264|title = Neuroelectronics: Smart connections|journal = Nature|volume = 503|issue = 7474|pages = 22–4|year = 2013|last1 = Waldrop|first1 = M. Mitchell|bibcode = 2013Natur.503...22W|doi-access = free}}</ref><ref>{{cite journal|doi=10.1109/JPROC.2014.2313565|title = Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations|journal = Proceedings of the IEEE|volume = 102|issue = 5|pages = 699–716|year = 2014|last1 = Benjamin|first1 = Ben Varkey|last2 = Peiran Gao|last3 = McQuinn|first3 = Emmett|last4 = Choudhary|first4 = Swadesh|last5 = Chandrasekaran|first5 = Anand R.|last6 = Bussat|first6 = Jean-Marie|last7 = Alvarez-Icaza|first7 = Rodrigo|last8 = Arthur|first8 = John V.|last9 = Merolla|first9 = Paul A.|last10 = Boahen|first10 = Kwabena|s2cid = 17176371}}</ref>

Line 94 ⟶ 95:

* [[Hardware for artificial intelligence]]

* [[Lithionics]]

* [[Neuromorphic Olfaction Systems]]

* [[Neurorobotics]]

* [[Optical flow sensor]]