--- title: Why Deep Learning Failed on Tables for a Decade - Frank... description: Frank Hutter discusses why deep learning historically underperformed on tabular data compared to domains like computer vision. He points to Google's TabNet... canonical: https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Why Deep Learning Failed on Tables for a Decade - Frank Hutter | daily.dev og:description: Frank Hutter discusses why deep learning historically underperformed on tabular data compared to domains like computer vision. He points to Google's TabNet... og:url: https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca og:image: https://api.daily.dev/og/posts/vbK2PuoCa.png og:image:alt: Why Deep Learning Failed on Tables for a Decade - Frank Hutter og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Why Deep Learning Failed on Tables for a Decade - Frank Hutter **[Machine Learning Street Talk](https://daily.dev/sources/ml_street_talk)** · 1 min read · 0 upvotes · 0 comments ## Summary Frank Hutter discusses why deep learning historically underperformed on tabular data compared to domains like computer vision. He points to Google's TabNet (2019) as a heavily cited but ultimately ineffective approach that failed to generalize across datasets, since tables from different domains (e.g., medicine vs. insurance) share no row-level structure to learn from. He argues the real breakthrough came from in-context learning, which allows transfer across datasets at the level of patterns rather than raw data, finally enabling deep learning to work well for tabular problems. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why did Google's TabNet fail to generalize well across different tabular datasets? TabNet, introduced by Google in 2019, struggled because tabular datasets from different domains share nothing at the row level to learn from transferably. Unlike image data, where one large diverse dataset like ImageNet lets models learn generalizable visual patterns, a medical table and an insurance table have no comparable row-level structure, so the model cannot transfer knowledge between them despite the approach's popularity and high citation count. _Anyone evaluating tabular deep learning approaches can track breakthroughs like this by following ML research on daily.dev._ ### How does in-context learning enable deep learning to generalize across different tabular datasets? In-context learning allows models to transfer knowledge across different tabular datasets by operating on the level of patterns rather than individual rows or raw data values. Since row-level data from one domain, like medicine, tells you nothing about another domain, like insurance, the transfer instead happens at a more abstract pattern level, which is what finally made deep learning work well for tabular data. _Developers building tabular ML pipelines can keep up with in-context learning advances via daily.dev._ ## Similar posts on daily.dev - [Why Large Language Models Fail at Tabular Prediction](https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm) · Hacker News · 1 upvotes · 1 comments - [How the Rise of Tabular Foundation Models Is Reshaping Data Science](https://daily.dev/posts/how-the-rise-of-tabular-foundation-models-is-reshaping-data-science-2taes3vse) · Towards Data Science · 0 upvotes · 0 comments - [H2O.ai launches tabH2O foundation model for tabular data](https://daily.dev/posts/h2o-ai-launches-tabh2o-foundation-model-for-tabular-data-hwllsnijd) · The Next Web · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#deep-learning](https://daily.dev/tags/deep-learning) [View this post on daily.dev](https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Why Deep Learning Failed on Tables for a Decade - Frank Hutter","url":"https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca"},"datePublished":"2026-09-29T10:09:06.062Z","dateModified":"2026-09-29T10:09:24.483Z","description":"Frank Hutter discusses why deep learning historically underperformed on tabular data compared to domains like computer vision. He points to Google's TabNet...","image":"https://i.ytimg.com/vi/tQ3nMI-anxU/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/tQ3nMI-anxU/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Machine Learning Street Talk","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Machine Learning Street Talk","logo":"https://media.daily.dev/image/upload/s--dnAejGaH--/f_auto/v1732810141/logos/ml_street_talk","url":"https://daily.dev/sources/ml_street_talk"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,deep-learning","timeRequired":"PT1M","video":{"@type":"VideoObject","name":"Why Deep Learning Failed on Tables for a Decade - Frank Hutter","description":"Frank Hutter discusses why deep learning historically underperformed on tabular data compared to domains like computer vision. He points to Google's TabNet...","thumbnailUrl":"https://i.ytimg.com/vi/tQ3nMI-anxU/sddefault.jpg","uploadDate":"2026-09-29T10:09:06.062Z","duration":"PT1M","url":"https://api.daily.dev/r/vbK2PuoCa","embedUrl":"https://www.youtube.com/embed/tQ3nMI-anxU"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Machine Learning Street Talk","item":"https://daily.dev/sources/ml_street_talk"},{"@type":"ListItem","position":3,"name":"Why Deep Learning Failed on Tables for a Decade - Frank Hutter"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/why-deep-learning-failed-on-tables-for-a-decade---frank-hutter-vbk2puoca#faq","mainEntity":[{"@type":"Question","name":"Why did Google's TabNet fail to generalize well across different tabular datasets?","acceptedAnswer":{"@type":"Answer","text":"TabNet, introduced by Google in 2019, struggled because tabular datasets from different domains share nothing at the row level to learn from transferably. Unlike image data, where one large diverse dataset like ImageNet lets models learn generalizable visual patterns, a medical table and an insurance table have no comparable row-level structure, so the model cannot transfer knowledge between them despite the approach's popularity and high citation count. Anyone evaluating tabular deep learning approaches can track breakthroughs like this by following ML research on daily.dev."}},{"@type":"Question","name":"How does in-context learning enable deep learning to generalize across different tabular datasets?","acceptedAnswer":{"@type":"Answer","text":"In-context learning allows models to transfer knowledge across different tabular datasets by operating on the level of patterns rather than individual rows or raw data values. Since row-level data from one domain, like medicine, tells you nothing about another domain, like insurance, the transfer instead happens at a more abstract pattern level, which is what finally made deep learning work well for tabular data. Developers building tabular ML pipelines can keep up with in-context learning advances via daily.dev."}}]} ```