en.m.wikipedia.org
Connectionism: Difference between revisions
← Previous edit
Next edit →
Connectionism (edit)
Revision as of 13:24, 19 December 2019
332 BYTES REMOVED
,  1 YEAR AGO
m
Alter: url. Add: year, class, eprint, chapter-url, hdl, bibcode, pmid, pages, doi, issue. Removed or converted URL. Removed parameters. Formatted dashes. Some additions/deletions were actually parameter name changes. | You can use this bot yourself. Report bugs here.| Activated by User:Headbomb | via #UCB_webform
{{more footnotes|date=April 2014}}
'''Connectionism''' is an approach in the fields of [[cognitive science]] that hopes to explain [[mind|mental]] phenomena using [[artificial neural networks]] (ANN).<ref name=":0">{{cite book|url=https://plato.stanford.edu/archives/fall2018/entries/connectionism/|title=The Stanford Encyclopedia of Philosophy|first=James|last=Garson|editor-first=Edward N.|editor-last=Zalta|date=27 November 2018|publisher=Metaphysics Research Lab, Stanford University|via=Stanford Encyclopedia of Philosophy}}</ref> Connectionism presents a cognitive theory based on simultaneously occurring, distributed signal activity via connections that can be represented numerically, where learning occurs by modifying connection strengths based on experience.<ref name=":1">{{Cite journal|last=Smolensky|first=Paul|date=1999|title=Grammar-based Connectionist Approaches to Language|url=http://csjarchive.cogsci.rpi.edu/1999v23/i04/p0589p0613/MAIN.PDF|journal=Cognitive Science|volume=23​|issue=4​|pages=​589-613​589–613​|via=​|doi=10.1207/s15516709cog2304_9​}}</ref> Some advantages of the connectionist approach include its applicability to a broad array of functions, structural approximation to biological neurons, low requirements for innate structure, and capacity for [[graceful degradation]].<ref name=":2">{{Cite book|title=The Algebraic Mind: Integrating Connectionism and Cognitive Science (Learning, Development, and Conceptual Change)|last=Marcus|first=Gary F.|publisher=MIT Press|year=2001|isbn=978-0262632683|location=Cambridge, Massachusetts|pages=​27-28​27–28​}}</ref> Some disadvantages include the difficulty in deciphering how ANNs process information and a resultant difficulty explaining phenomena at a higher level.<ref name=":1" /> The success of [[deep learning]] networks in the past decade has greatly increased the popularity of this approach, but the complexity and scale of such networks has brought with them increased [[Explainable artificial intelligence|interpretability problems]].<ref name=":0" /> Connectionism is seen by many to offer an alternative to classical theories of mind based on symbolic computation, but the extent to which the two approaches are compatible has been the subject of much debate since their inception.<ref name=":0" />
[[File:Artificial_neural_network.svg|thumb|Connectionist (ANN) model with a hidden layer]]
 
 
===Biological realism===
Connectionist work in general does not need to be biologically realistic and therefore suffers from a lack of neuroscientific plausibility.<ref>{{Cite web|url=http://www.encephalos.gr/48-1-01e.htm|title=Encephalos Journal|website=www.encephalos.gr|access-date=2018-02-20}}</ref><ref>{{Cite book|url=https://books.google.com/​books​?id=s-OCCwAAQBAJ&pg=PT18&lpg=PT18&dq=%22accurate%22​&f=false​#v=onepage&q=%22accurate%22&f=false|title=Neural Geographies: Feminism and the Microstructure of Cognition|last=Wilson|first=Elizabeth A.|date=2016-02-04|publisher=Routledge|isbn=9781317958765|language=en}}</ref><ref>{{Cite web|url=https://pdfs.semanticscholar.org/e953/59bc80e624a963a3d8c943e3b2898a397ef7.pdf|title=Organismically-inspired robotics: homeostatic adaptation and teleology beyond the closed sensorimotor loop|last=|first=|date=|website=|access-date=}}</ref><ref>{{Cite journal|last=Zorzi|first=Marco|last2=Testolin|first2=Alberto|last3=Stoianov|first3=Ivilin P.|date=2013-08-20|title=Modeling language and cognition with deep unsupervised learning: a tutorial overview|journal=Frontiers in Psychology|volume=4​|pages=515​|doi=10.3389/fpsyg.2013.00515|issn=1664-1078|pmc=3747356|pmid=23970869}}</ref><ref>{{Cite web|url=http://scholarworks.sjsu.edu/cgi/viewcontent.cgi?article=1015&context=comparativephilosophy|title=ANALYTIC AND CONTINENTAL PHILOSOPHY|last=|first=|date=|website=|access-date=}}</ref><ref>{{Cite book|url=https://books.google.com/​books​?id=uV9TZzOITMwC&pg=PA17&lpg=PA17&dq=%22biological%20plausibility%22​&f=false​#v=onepage&q=%22biological%20plausibility%22&f=false|title=Neural Network Perspectives on Cognition and Adaptive Robotics|last=Browne|first=A.|date=1997-01-01|publisher=CRC Press|isbn=9780750304559|language=en}}</ref><ref>{{Cite book|url=https://books.google.com/books?id=7pPv0STSos8C&pg=PA63&lpg=PA63&dq=%22biological+realism%22#v=onepage&q=%22biological%20realism%22&f=false|title=Connectionism in Perspective|last=Pfeifer|first=R.|last2=Schreter|first2=Z.|last3=Fogelman-Soulié|first3=F.|last4=Steels|first4=L.|date=1989-08-23|publisher=Elsevier​|year=​|isbn=9780444598769|location=|pages=|language=en}}</ref> However, the structure of neural networks is derived from that of biological [[Neuron|neurons]], and this parallel in low-level structure is often argued to be an advantage of connectionism in modeling cognitive structures compared with other approaches.<ref name=":2" /> One area where connectionist models are thought to be biologically implausible is with respect to error-propagation networks that are needed to support learning <ref>{{Cite journal|last=Crick|first=Francis|date=1989-01|title=The recent excitement about neural networks​|url=https://www.nature.com/articles/337129a0​|journal=Nature|language=en|volume=337|issue=6203|pages=129–132|doi=10.1038/337129a0​|pmid=2911347​|issn=1476-4687​|bibcode=1989Natur.337..129C​}}</ref><ref name=":4">{{Cite journal|last=Rumelhart|first=David E.|last2=Hinton|first2=Geoffrey E.|last3=Williams|first3=Ronald J.|date=1986-10|title=Learning representations by back-propagating errors​|url=https://www.nature.com/articles/323533a0​|journal=Nature|language=en|volume=323|issue=6088|pages=533–536|doi=10.1038/323533a0|issn=1476-4687​|bibcode=1986Natur.323..533R​}}</ref>, but error propagation can explain some of the biologically-generated electrical activity seen at the scalp in [[Event-related potential|event-related potentials]] such as the [[N400 (neuroscience)|N400]] and [[P600 (neuroscience)|P600]] <ref>{{Cite journal|last=Fitz|first=Hartmut|last2=Chang|first2=Franklin|date=2019-06-01|title=Language ERPs reflect learning through prediction error propagation​|url=http://www.sciencedirect.com/science/article/pii/S0010028518300124​|journal=Cognitive Psychology|volume=111|pages=15–52|doi=10.1016/j.cogpsych.2019.03.002|issn=0010-0285​|hdl=21.11116/0000-0003-474F-6​}}</ref>, and this provides some biological support for one of the key assumptions of connectionist learning procedures.
 
===Learning===
The weights in a neural network are adjusted according to some [[learning rule]] or algorithm, such as [[Hebbian theory|Hebbian learning]]. Thus, connectionists have created many sophisticated learning procedures for neural networks. Learning always involves modifying the connection weights. In general, these involve mathematical formulas to determine the change in weights when given sets of data consisting of activation vectors for some subset of the neural units. Several studies have been focused on designing teaching-learning methods based on connectionism.<ref>{{Cite journal|last=Novo|first=María-Luisa|last2=Alsina|first2=Ángel|last3=Marbán|first3=José-María|last4=Berciano|first4=Ainhoa|date=2017|title=Connective Intelligence for Childhood Mathematics Education​|url=https://doi.org/10.3916/C52-2017-03​|journal=Comunicar|language=es|volume=25|issue=52|pages=29–39|doi=10.3916/c52-2017-03|issn=1134-3478}}</ref>
 
By formalizing learning in such a way, connectionists have many tools. A very common strategy in connectionist learning methods is to incorporate [[gradient descent]] over an error surface in a space defined by the weight matrix. All gradient descent learning in connectionist models involves changing each weight by the [[partial derivative]] of the error surface with respect to the weight. [[Backpropagation]] (BP), first made popular in the 1980s, is probably the most commonly known connectionist gradient descent algorithm today <ref name=":4" />.
Though PDP is the dominant form of connectionism, other theoretical work should also be classified as connectionist.
 
Many connectionist principles can be traced to early work in [[psychology]], such as that of [[William James]].<ref>{{cite book |last1=Anderson |first1= James A.|last2=Rosenfeld |first2= Edward |date= 1989|title= Neurocomputing: Foundations of Research|​chapter-​url= |location= |publisher= A Bradford Book |page= 1|chapter = Chapter 1: (1890) William James ''Psychology (Brief Course)'' |isbn=978-0262510486 |accessdate= }}</ref> Psychological theories based on knowledge about the human brain were fashionable in the late 19th century. As early as 1869, the neurologist [[John Hughlings Jackson]] argued for multi-level, distributed systems. Following from this lead, [[Herbert Spencer]]'s ''Principles of Psychology'', 3rd edition (1872), and [[Sigmund Freud]]'s ''Project for a Scientific Psychology'' (composed 1895) propounded connectionist or proto-connectionist theories. These tended to be speculative theories. But by the early 20th century, [[Edward Thorndike]] was experimenting on learning that posited a connectionist type network.
 
[[Friedrich Hayek]] independently conceived the Hebbian synapse learning model in a paper presented in 1920 and developed that model into global brain theory constituted of networks Hebbian synapses building into larger systems of maps and memory network {{Citation needed|date=March 2015}}. Hayek's breakthrough work was cited by Frank Rosenblatt in his perceptron paper.
* Computationalists often posit [[domain specificity|domain specific]] symbolic sub-systems designed to support learning in specific areas of cognition (e.g., language, intentionality, number), whereas connectionists posit one or a small set of very general learning-mechanisms.
 
Despite these differences, some theorists have proposed that the connectionist architecture is simply the manner in which organic brains happen to implement the symbol-manipulation system. This is logically possible, as it is well known that connectionist models can implement symbol-manipulation systems of the kind used in computationalist models<ref name=":3">{{Cite journal|last=Chang|first=Franklin|date=2002|title=Symbolically speaking: a connectionist model of sentence production​|url=https://onlinelibrary.wiley.com/doi/abs/10.1207/s15516709cog2605_3​|journal=Cognitive Science|language=en|volume=26|issue=5|pages=609–651|doi=10.1207/s15516709cog2605_3|issn=1551-6709}}</ref>, as indeed they must be able if they are to explain the human ability to perform symbol-manipulation tasks. Several cognitive models combining both symbol-manipulative and connectionist architectures have been proposed, notably among them [[Paul Smolensky]]'s Integrated Connectionist/Symbolic Cognitive Architecture (ICS).<ref name=":0" /><ref>{{Cite journal|last=Smolensky|first=Paul|date=1990|title=Tensor Product Variable Binding and the Representation of Symbolic Structures in Connectionist Systems|url=http://www.lscp.net/persons/dupoux/teaching/AT1_2012/papers/Smolensky_1990_TensorProductVariableBinding.AI.pdf|journal=Artificial Intelligence|volume=46​|issue=1–2​|pages=​159-216​159–216​|via=​|doi=10.1016/0004-3702(90)90007-M​}}</ref> But the debate rests on whether this symbol manipulation forms the foundation of cognition in general, so this is not a potential vindication of computationalism. Nonetheless, computational descriptions may be helpful high-level descriptions of cognition of logic, for example.
 
The debate was largely centred on logical arguments about whether connectionist networks could produce the syntactic structure observed in this sort of reasoning. This was later achieved although using fast-variable binding abilities outside of those standardly assumed in connectionist models<ref name=":3" /><ref>{{Cite journal|last=Shastri|first=Lokendra|last2=Ajjanagadde|first2=Venkat|date=1993/09|title=From simple associations to systematic reasoning: A connectionist representation of rules, variables and dynamic bindings using temporal synchrony​|url=https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/from-simple-associations-to-systematic-reasoning-a-connectionist-representation-of-rules-variables-and-dynamic-bindings-using-temporal-synchrony/A00CC41AFE06B361E644DAC5ED8F65A3​|journal=Behavioral and Brain Sciences|language=en|volume=16|issue=3|pages=417–451|doi=10.1017/S0140525X00030910|issn=1469-1825}}</ref>. {{As of | 2016}}, progress in neurophysiology and general advances in the understanding of neural networks have led to the successful modelling of a great many of these early problems, and the debate about fundamental cognition has, thus, largely been decided among neuroscientists in favour of connectionism.{{citation needed|date=March 2012}} However, these fairly recent developments have yet to reach consensus acceptance among those working in other fields, such as psychology or philosophy of mind.
 
Part of the appeal of computational descriptions is that they are relatively easy to interpret, and thus may be seen as contributing to our understanding of particular mental processes, whereas connectionist models are in general more opaque, to the extent that they may be describable only in very general terms (such as specifying the learning algorithm, the number of units, etc.), or in unhelpfully low-level terms. In this sense connectionist models may instantiate, and thereby provide evidence for, a broad theory of cognition (i.e., connectionism), without representing a helpful theory of the particular process that is being modelled. In this sense the debate might be considered as to some extent reflecting a mere difference in the level of analysis in which particular theories are framed. Some researchers suggest that the analysis gap is the consequence of connectionist mechanisms giving rise to [[Emergence|emergent phenomena]] that may be describable in computational terms.<ref>{{Cite journal|last=Ellis|first=Nick C.|date=1998|title=Emergentism, Connectionism and Language Learning|url=http://www-personal.umich.edu/~ncellis/NickEllis/Publications_files/Emergentism.pdf|journal=Language Learning|volume=48:4|pages=​631-664​631–664​|via=}}</ref>
 
The recent{{when|date=February 2016}} popularity of [[Cognitive Model#Dynamical systems|dynamical systems]] in [[philosophy of mind]] have added a new perspective on the debate; some authors{{which|date=February 2016}} now argue that any split between connectionism and computationalism is more conclusively characterized as a split between computationalism and [[Cognitive Model#Dynamical systems|dynamical systems]].
 
In 2014, [[Alex Graves (computer scientist)|Alex Graves]] and others from [[DeepMind]] published a series of papers describing a novel Deep Neural Network structure called the [[Neural Turing Machine]]<ref>{{cite arxiv|last1=Graves|first1=Alex|title=Neural Turing Machines|​arxiv​eprint​=1410.5401​|class=cs.NE|year=2014​}}</ref> able to read symbols on a tape and store symbols in memory. Relational Networks, another Deep Network module published by DeepMind are able to create object-like representations and manipulate them to answer complex questions. Relational Networks and Neural Turing Machines are further evidence that connectionism and computationalism need not be at odds.
 
==See also==
Citation bot
BOTS
1,782,661
EDITS
Content is available under CC BY-SA 3.0 unless otherwise noted.
Privacy policy
Terms of Use
Desktop
HomeRandomNearbyLog inSettingsDonateAbout WikipediaDisclaimers