{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,6]],"date-time":"2026-06-06T20:02:38Z","timestamp":1780776158280,"version":"3.54.1"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1012286","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,2,20]],"date-time":"2025-02-20T00:00:00Z","timestamp":1740009600000}}],"reference-count":69,"publisher":"Public Library of Science (PLoS)","issue":"2","license":[{"start":{"date-parts":[[2025,2,10]],"date-time":"2025-02-10T00:00:00Z","timestamp":1739145600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>\n                    There is an important challenge in systematically interpreting the internal representations of deep neural networks (DNNs). Existing techniques are often less effective for non-tabular tasks, or they primarily focus on qualitative, ad-hoc interpretations of models. In response, this study introduces a cognitive science-inspired, multi-dimensional quantification and visualization approach that captures two temporal dimensions of model learning: the \u201cinformation-processing trajectory\u201d and the \u201cdevelopmental trajectory.\u201d The former represents the influence of incoming signals on an agent\u2019s decision-making, while the latter conceptualizes the gradual improvement in an agent\u2019s performance throughout its lifespan. Tracking the learning curves of DNNs enables researchers to explicitly identify the model appropriateness of a given task, examine the properties of the underlying input signals, and assess the model\u2019s alignment (or lack thereof) with human learning experiences. To illustrate this method, we conducted 750 runs of simulations on two temporal tasks: gesture detection and sentence classification, showcasing its applicability across different types of deep learning tasks. Using four descriptive metrics to quantify the mapped learning curves\u2014\n                    <jats:italic>start<\/jats:italic>\n                    ,\n                    <jats:italic>end - start<\/jats:italic>\n                    ,\n                    <jats:italic>max<\/jats:italic>\n                    ,\n                    <jats:italic>\n                      t\n                      <jats:sub>max<\/jats:sub>\n                    <\/jats:italic>\n                    \u2014, we identified significant differences in learning patterns based on data sources and class distinctions (all\n                    <jats:italic>p\u2019s<\/jats:italic>\n                    \u00a0&lt;\u00a0 .0001), the prominent role of spatial semantics in gesture learning, and larger information gains in language learning. We highlight three key insights gained from mapping learning curves:\n                    <jats:italic>non-monotonic progress<\/jats:italic>\n                    ,\n                    <jats:italic>pairwise comparisons<\/jats:italic>\n                    , and\n                    <jats:italic>domain distinctions<\/jats:italic>\n                    . We reflect on the theoretical implications of this method for cognitive processing, language models and representations from multiple modalities.\n                  <\/jats:p>","DOI":"10.1371\/journal.pcbi.1012286","type":"journal-article","created":{"date-parts":[[2025,2,10]],"date-time":"2025-02-10T13:42:08Z","timestamp":1739194928000},"page":"e1012286","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":4,"title":["Mapping the learning curves of deep learning networks"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-9308-2310","authenticated-orcid":true,"given":"Yanru","family":"Jiang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rick","family":"Dale","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2025,2,10]]},"reference":[{"issue":"7553","key":"pcbi.1012286.ref001","doi-asserted-by":"crossref","first-page":"436","DOI":"10.1038\/nature14539","article-title":"Deep learning","volume":"521","author":"Y LeCun","year":"2015","journal-title":"Nature"},{"key":"pcbi.1012286.ref002"},{"key":"pcbi.1012286.ref003"},{"key":"pcbi.1012286.ref004"},{"key":"pcbi.1012286.ref005","volume-title":"Curran Associates, Inc.","author":"A Morcos","year":"2018"},{"key":"pcbi.1012286.ref006"},{"issue":"6","key":"pcbi.1012286.ref007","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3457607","article-title":"A survey on bias and fairness in machine learning","volume":"54","author":"N Mehrabi","year":"2021","journal-title":"ACM Comput Surv"},{"issue":"12","key":"pcbi.1012286.ref008","doi-asserted-by":"crossref","first-page":"5826","DOI":"10.3390\/app12125826","article-title":"Discrimination, bias, fairness, and trustworthy AI","volume":"12","author":"D Varona","year":"2022","journal-title":"Appl Sci"},{"issue":"23","key":"pcbi.1012286.ref009","doi-asserted-by":"crossref","first-page":"8619","DOI":"10.1073\/pnas.1403112111","article-title":"Performance-optimized hierarchical models predict neural responses in higher visual cortex","volume":"111","author":"DLK Yamins","year":"2014","journal-title":"Proc Natl Acad Sci U S A"},{"issue":"3","key":"pcbi.1012286.ref010","doi-asserted-by":"crossref","first-page":"201","DOI":"10.1037\/0003-066X.33.3.201","article-title":"Managing motivation to expand human freedom","volume":"33","author":"DC McClelland","year":"1978","journal-title":"Am Psychol"},{"issue":"3","key":"pcbi.1012286.ref011","doi-asserted-by":"crossref","first-page":"369","DOI":"10.1038\/s41593-022-01026-4","article-title":"Shared computational principles for language processing in humans and deep language models","volume":"25","author":"A Goldstein","year":"2022","journal-title":"Nat Neurosci"},{"issue":"45","key":"pcbi.1012286.ref012","doi-asserted-by":"crossref","first-page":"e2105646118","DOI":"10.1073\/pnas.2105646118","article-title":"The neural architecture of language: Integrative modeling converges on predictive processing","volume":"118","author":"M Schrimpf","year":"2021","journal-title":"Proc Natl Acad Sci U S A"},{"issue":"2","key":"pcbi.1012286.ref013","doi-asserted-by":"crossref","first-page":"179","DOI":"10.1207\/s15516709cog1402_1","article-title":"Finding structure in time","volume":"14","author":"JL Elman","year":"1990","journal-title":"Cognit Sci"},{"key":"pcbi.1012286.ref014"},{"issue":"4","key":"pcbi.1012286.ref015","doi-asserted-by":"crossref","first-page":"547","DOI":"10.1111\/j.1551-6709.2009.01023.x","article-title":"On the meaning of words and dinosaur bones: lexical knowledge without a lexicon","volume":"33","author":"JL Elman","year":"2009","journal-title":"Cogn Sci"},{"issue":"5598","key":"pcbi.1012286.ref016","doi-asserted-by":"crossref","first-page":"1569","DOI":"10.1126\/science.298.5598.1569","article-title":"The faculty of language: what is it, who has it, and how did it evolve?","volume":"298","author":"MD Hauser","year":"2002","journal-title":"Science"},{"key":"pcbi.1012286.ref017"},{"issue":"1","key":"pcbi.1012286.ref018","doi-asserted-by":"crossref","first-page":"293","DOI":"10.1162\/coli_a_00492","article-title":"Language model behavior: a comprehensive survey","volume":"50","author":"TA Chang","year":"2024","journal-title":"Comput Linguist"},{"key":"pcbi.1012286.ref019"},{"key":"pcbi.1012286.ref020"},{"key":"pcbi.1012286.ref021"},{"key":"pcbi.1012286.ref022"},{"issue":"4","key":"pcbi.1012286.ref023","doi-asserted-by":"crossref","first-page":"310","DOI":"10.1038\/nrn1076","article-title":"The parallel distributed processing approach to semantic cognition","volume":"4","author":"JL McClelland","year":"2003","journal-title":"Nat Rev Neurosci"},{"issue":"3","key":"pcbi.1012286.ref024","doi-asserted-by":"crossref","first-page":"31","DOI":"10.1145\/3236386.3241340","article-title":"The Mythos of model interpretability","volume":"16","author":"ZC Lipton","year":"2018","journal-title":"Queue"},{"key":"pcbi.1012286.ref025"},{"key":"pcbi.1012286.ref026"},{"key":"pcbi.1012286.ref027"},{"issue":"7600","key":"pcbi.1012286.ref028","doi-asserted-by":"crossref","first-page":"453","DOI":"10.1038\/nature17637","article-title":"Natural speech reveals the semantic maps that tile human cerebral cortex","volume":"532","author":"AG Huth","year":"2016","journal-title":"Nature"},{"issue":"1","key":"pcbi.1012286.ref029","doi-asserted-by":"crossref","first-page":"4309","DOI":"10.1038\/s41467-023-39872-w","article-title":"Phonemic segmentation of narrative speech in human cerebral cortex","volume":"14","author":"XL Gong","year":"2023","journal-title":"Nat Commun"},{"issue":"6","key":"pcbi.1012286.ref030","doi-asserted-by":"crossref","first-page":"1210","DOI":"10.1016\/j.neuron.2012.10.014","article-title":"A continuous semantic space describes the representation of thousands of object and action categories across the human brain","volume":"76","author":"AG Huth","year":"2012","journal-title":"Neuron"},{"issue":"2","key":"pcbi.1012286.ref031","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3436494","article-title":"Visual semantic-based representation learning using deep CNNs for scene recognition","volume":"17","author":"S Gupta","year":"2021","journal-title":"ACM Trans Multim Comput Commun Appl"},{"key":"pcbi.1012286.ref032","doi-asserted-by":"crossref","first-page":"107470","DOI":"10.1016\/j.knosys.2021.107470","article-title":"Content and context features for scene image representation","volume":"232","author":"C Sitaula","year":"2021","journal-title":"Knowl-Based Syst"},{"issue":"10","key":"pcbi.1012286.ref033","doi-asserted-by":"crossref","first-page":"e0223792","DOI":"10.1371\/journal.pone.0223792","article-title":"THINGS: a database of 1,854 object concepts and more than 26,000 naturalistic object images","volume":"14","author":"MN Hebart","year":"2019","journal-title":"PLoS One"},{"key":"pcbi.1012286.ref034"},{"key":"pcbi.1012286.ref035","first-page":"101184","article-title":"Transformability, generalizability, but limited diffusibility: comparing global vs","volume":"83","author":"Y Jiang","year":"2024","journal-title":"task-specific language representations in deep neural networks. Cognit Syst Res"},{"issue":"3","key":"pcbi.1012286.ref036","doi-asserted-by":"crossref","first-page":"416","DOI":"10.1016\/j.neuron.2019.12.002","article-title":"Direct fit to nature: an evolutionary perspective on biological and artificial neural networks","volume":"105","author":"U Hasson","year":"2020","journal-title":"Neuron"},{"issue":"7","key":"pcbi.1012286.ref037","doi-asserted-by":"crossref","first-page":"1025","DOI":"10.1038\/nn.4042","article-title":"A neural network that finds a naturalistic solution for the production of muscle activity","volume":"18","author":"D Sussillo","year":"2015","journal-title":"Nat Neurosci"},{"key":"pcbi.1012286.ref038"},{"key":"pcbi.1012286.ref039"},{"issue":"7","key":"pcbi.1012286.ref040","doi-asserted-by":"crossref","first-page":"e13312","DOI":"10.1111\/cogs.13312","article-title":"Modeling structure-building in the brain with CCG parsing and large language models","volume":"47","author":"M Stanojevi\u0107","year":"2023","journal-title":"Cogn Sci"},{"issue":"2","key":"pcbi.1012286.ref041","doi-asserted-by":"crossref","first-page":"334","DOI":"10.1016\/j.fcij.2018.10.003","article-title":"Time series forecasting using artificial neural networks methodologies: a systematic review","volume":"3","author":"A Tealab","year":"2018","journal-title":"Future Comput Inf J"},{"key":"pcbi.1012286.ref042"},{"key":"pcbi.1012286.ref043"},{"key":"pcbi.1012286.ref044"},{"issue":"10","key":"pcbi.1012286.ref045","doi-asserted-by":"crossref","first-page":"603","DOI":"10.1016\/j.tics.2015.07.013","article-title":"Arbitrariness, iconicity, and systematicity in language","volume":"19","author":"M Dingemanse","year":"2015","journal-title":"Trends Cogn Sci"},{"issue":"11","key":"pcbi.1012286.ref046","doi-asserted-by":"crossref","first-page":"456","DOI":"10.1016\/S1364-6613(02)01990-3","article-title":"The past and future of the past tense","volume":"6","author":"S Pinker","year":"2002","journal-title":"Trends Cogn Sci"},{"key":"pcbi.1012286.ref047","unstructured":"Joe Shishido. Joe Shishido gesture 2020-01-21.gif. January 21, 2020. Licensed under Creative Commons. [cited 2024 Nov 02]. Available from: https:\/\/commons.wikimedia.org\/wiki\/File:Joe_Shishido_gesture_2020-01-21.gif"},{"key":"pcbi.1012286.ref048"},{"key":"pcbi.1012286.ref049"},{"issue":"4","key":"pcbi.1012286.ref050","first-page":"344","article-title":"The nature of emotions: human emotions have deep evolutionary roots","volume":"89","author":"R Plutchik","year":"2001","journal-title":"a fact that may explain their complexity and provide tools for clinical practice. Am Scientist."},{"key":"pcbi.1012286.ref051"},{"key":"pcbi.1012286.ref052","article-title":"Twitter sentiment classification using distant supervision","author":"A Go","year":"2009"},{"key":"pcbi.1012286.ref053","article-title":"Efficient estimation of word representations in vector space. In: International Conference on Learning Representations","author":"T Mikolov","year":"2013"},{"key":"pcbi.1012286.ref054"},{"issue":"3","key":"pcbi.1012286.ref055","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/2682899","article-title":"A review and meta-analysis of multimodal affect detection systems","volume":"47","author":"SK D\u2019mello","year":"2015","journal-title":"ACM Comput Surv"},{"key":"pcbi.1012286.ref056"},{"key":"pcbi.1012286.ref057"},{"key":"pcbi.1012286.ref058"},{"key":"pcbi.1012286.ref059","article-title":"Automated nonverbal cue detection in political-debate videos: an optimized RNN-LSTM approach. In: International Conference on Human-Computer Interaction. Springer","author":"Y Jiang","year":"2023"},{"issue":"2","key":"pcbi.1012286.ref060","doi-asserted-by":"crossref","first-page":"1883","DOI":"10.4249\/scholarpedia.1883","article-title":"K-nearest neighbor","volume":"4","author":"L Peterson","year":"2009","journal-title":"Scholarpedia"},{"key":"pcbi.1012286.ref061","article-title":"KNN vs SVM: a comparison of algorithms","author":"RG Pacheco","year":"2017"},{"issue":"1","key":"pcbi.1012286.ref062","doi-asserted-by":"crossref","first-page":"137","DOI":"10.1207\/s15327647jcd0501_14","article-title":"U-shaped curves in development: a PDP approach","volume":"5","author":"TT Rogers","year":"2004","journal-title":"J Cognit Develop"},{"key":"pcbi.1012286.ref063"},{"key":"pcbi.1012286.ref064","article-title":"SVCCA: singular vector canonical correlation analysis for deep learning dynamics and interpretability. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. NIPS 2017. Red Hook, NY, USA: Curran Associates Inc.","author":"M Raghu","year":"2017"},{"issue":"1","key":"pcbi.1012286.ref065","doi-asserted-by":"crossref","first-page":"5725","DOI":"10.1038\/s41467-020-19632-w","article-title":"Individual differences among deep neural network models","volume":"11","author":"J Mehrer","year":"2020","journal-title":"Nat Commun"},{"key":"pcbi.1012286.ref066"},{"key":"pcbi.1012286.ref067","unstructured":"Lundberg SM, Lee S. A unified approach to interpreting model predictions. In: Neural Information Processing Systems. 2017. Available from: https:\/\/api.semanticscholar.org\/CorpusID:21889700"},{"key":"pcbi.1012286.ref068"},{"issue":"13","key":"pcbi.1012286.ref069","doi-asserted-by":"crossref","first-page":"3521","DOI":"10.1073\/pnas.1611835114","article-title":"Overcoming catastrophic forgetting in neural networks","volume":"114","author":"J Kirkpatrick","year":"2017","journal-title":"Proc Natl Acad Sci U S A"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1012286","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,2,20]],"date-time":"2025-02-20T00:00:00Z","timestamp":1740009600000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012286","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,2,20]],"date-time":"2025-02-20T13:39:50Z","timestamp":1740058790000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012286"}},"subtitle":[],"editor":[{"given":"Varun","family":"Dutt","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2025,2,10]]},"references-count":69,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2025,2,10]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1012286","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2024.07.01.601491","asserted-by":"object"}]},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,2,10]]}}}