[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-42854-en":3,"doc-seo-42854-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},42854,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1782698725881665579",8,"Research & Report","Connectionist Modeling in Cognitive Science","Connectionist modeling in cognitive science explains how neural networks perform cognitive tasks through distributed, robust processing and continuous input–output mappings. The material outlines core neural-network properties, contrasts symbolic and connectionist approaches, and introduces major machine-learning paradigms. It details training procedures for discrete and continuous perceptrons, including perceptron learning rules and convergence for linearly separable classes. It further develops error concepts, two-layer perceptrons, and back-propagation, including forward/backward passes and weight-update equations.","|  |  |  |\n| --- | --- | --- |\n| Connectionist modeling in cognitive science\u003Cbr>Igor Farkaš\u003Cbr>Centre for Cognitive Science / Department of Applied Informatics Faculty of Mathematics, Physics and Informatics Comenius University in Bratislava\u003Cbr>[http://cogsci.fmph.uniba.sk/](http://cogsci.fmph.uniba.sk/)~farkas\u003Cbr>MEi:CogSci joint teaching, 9.11.2020, University of Vienna |  |  |\n\n\n| Modeling frameworks |  |  |\n| --- | --- | --- |\n| Symbolic 1950-\u003Cbr>\u003Cbr>Dynamic 1990-\u003Cbr> | Connectionist 1970-\u003Cbr>\u003Cbr>Probabilistic 2000-\u003Cbr> | 2 |\n| Farkaš (Journal of Cognitive Science, 2012) |  |  |\n\nCore properties of neural networks  \n● Neurobiological inspiration with abstraction  \n● Simple computing elements (neurons)  \n● Nonlinear interactions between them (excitation, inhibition)  \n● Ability to learn (elementary changes to knowledge representation)  \n● … based on experience (training examples)  \n● Distributedness and robustness  \n● Realisation of (continuous) input–output mappings  \n● … in Euclidean space  \n● … in case of recurrent models, with context dependency  \n3  \n\n| Principle ways of machine learning\u003Cbr>supervised (with teacher) unsupervised (self-organized) |  |\n| --- | --- |\n|  |  |\n| reinforcement learning (partial feedback)\u003Cbr>4\u003Cbr> |  |\n\nTraining of neural nets  \n● Like humans, NNs learn from experience (via training examples), rather than receive knowledge from the designer (the typical case of symbolic systems) .  \n● Each step (iteration) comprises these steps: (1) input patternselection, (2) output generation, (3) parameter (weights) update.  \n● Network performance is tested (only steps 1 and 2) using novel examples (inputs) to check that it can generalize from knowledge it acquired during training.  \n● Parameter update depends on the paradigm: (1) supervised, (2) unsupervised (self-organizing), (3) reinforcement learning.  \n● Neural nets require data representations in a numeric format.  \n5  \n\n| Supervised learning-discrete perceptron |  |\n| --- | --- |\n| ●\u003Cbr>●\u003Cbr>●\u003Cbr>●\u003Cbr>● | (Rosenblatt, 1962)\u003Cbr>Inputs x , weights w, output y Output:\u003Cbr>y = f(∑nj=1 wjxj- 􀀄)\u003Cbr>f = threshold function: unipolar {0,1} or bipolar {-1,+1}\u003Cbr>Supervised learning – uses teacher\u003Cbr>signal d unipolar f.\u003Cbr>y\u003Cbr>Learning rule: (with learning rate 􀀂)\u003Cbr>net\u003Cbr>wj(t+1) = wj(t) + 􀀂 (d – y) xj\u003Cbr>6\u003Cbr> |\n\nPerceptron algorithm  \nGiven: training data: input-target {x, d} pairs, unipolar perceptron Training:  \n0. Initialization: randomize weights, set learning rate, E = 0  \n1. choose input x , compute output y  \n2. compute instant. error e(t) = ½ (d – y)2, E = E + e(t)  \n3. if e(t) > 0, adjust weights using delta rule  \n4. if all patterns used, then goto 5, else go to 1  \n5. if E == 0 (all patterns in the set classified correctly), then end else  \nreorder inputs, E = 0 , go to 1  \n7  \n\n| Perceptron's separation capacity\u003Cbr>w1x1 + w2x2 +... + wnxn = 􀀄 linear separability of two classes\u003Cbr>[n](n-dim. case)[-dim. case](n-dim. case).  |  |  |\n| --- | --- | --- |\n| 2D\u003Cbr>example |  | linear decision boundary |\n| Fixed-increment convergence theorem (Rosenblatt, 1962): “Let the classes A and B are nite and linearly separable, then perceptron learning algorithm converges (updates its weight vector) in a nite number of steps.”\u003Cbr>8 |  |  |\n\nContinuous perceptron  \n● Nonlinear unit with sigmoid activ. function: y = f(net) = 1 / (1+e-net)  \n-has nice properties (boundedness, monotonicity, differentiability)  \n● Quadratic error function: e(w) = ½ (d – y)2  \n● Error minimization: we want e(w*) ≤ e(w)  \n● (stochastic, online) gradient descent learning:  \nwj(t+1) = wj(t) + 􀀂 (d(p)–y(p)) f '(net) xje(w)  \nsigmoid  \nf ’(net) = derivative of [f w.r.t. net](f w.r.t. net) → determines the function slope  \nContinuous perceptron as a binary classifier  \n9  \nw  \nw*  \nTwo-layer perceptron  \n● Inputs x , weights w, v, outputs y  \n● Nonlinear activation function f  \n● Unit activation:  \nhk = f 􀀆 ∑nj11 vkj xj 􀀉 yi = f 􀀆 ∑qk11 wik h k 􀀉  \n● Bias input: xn􀀈1 =hq 􀀈1=−1  \n● Activat","cbCait0tdUoi2pTi","https://ap.wps.com/l/cbCait0tdUoi2pTi","pdf",2668311,4,1,7,"English","en",105,"# Modeling frameworks\n## Core properties of neural networks\n# Principle ways of machine learning\n## Training of neural nets\n# Supervised learning: perceptron\n## Perceptron algorithm\n## Perceptron's separation capacity\n## Continuous perceptron\n## Two-layer perceptron\n# Error and back-propagation\n## Error surface and back-propagating error\n## Learning equations for original BP\n## Summary of back-propagation algorithm","[{\"question\":\"What core properties make neural networks suitable for connectionist modeling?\",\"answer\":\"Neural networks use neurobiological inspiration with abstraction, employ simple nonlinear units, support learning from training examples, and implement distributed representations that are robust. They realize continuous input–output mappings, including in recurrent settings with context dependence.\"},{\"question\":\"How does supervised perceptron learning update weights?\",\"answer\":\"The perceptron algorithm computes an output for an input pattern, measures instantaneous error relative to a teacher signal, and adjusts weights (via a delta rule) when the error is positive. Training iterates through patterns until all are classified correctly or a stopping condition is met.\"},{\"question\":\"How does back-propagation compute weight updates in multilayer networks?\",\"answer\":\"Back-propagation performs a forward pass to compute outputs and errors, then a backward pass to compute local unit errors at hidden and output layers. Weight changes follow the form of learning rate times unit error times unit input, iterating through the training set until a stopping criterion is satisfied.\"}]",1783372058,18,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"connectionist-modeling-in-cognitive-science","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/connectionist-modeling-in-cognitive-science/42854/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-20","2026-07-06",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What core properties make neural networks suitable for connectionist modeling?","Question",{"text":75,"@type":76},"Neural networks use neurobiological inspiration with abstraction, employ simple nonlinear units, support learning from training examples, and implement distributed representations that are robust. They realize continuous input–output mappings, including in recurrent settings with context dependence.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does supervised perceptron learning update weights?",{"text":80,"@type":76},"The perceptron algorithm computes an output for an input pattern, measures instantaneous error relative to a teacher signal, and adjusts weights (via a delta rule) when the error is positive. Training iterates through patterns until all are classified correctly or a stopping condition is met.",{"name":82,"@type":73,"acceptedAnswer":83},"How does back-propagation compute weight updates in multilayer networks?",{"text":84,"@type":76},"Back-propagation performs a forward pass to compute outputs and errors, then a backward pass to compute local unit errors at hidden and output layers. Weight changes follow the form of learning rate times unit error times unit input, iterating through the training set until a stopping criterion is satisfied.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]