--- title: "Thomas Unterthiner — author profile and papers" description: "Thomas Unterthiner: 2 papers on Elamur Research (2015–2020), h-index 29 — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/thomas-unterthiner/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/thomas-unterthiner/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Thomas Unterthiner # Thomas Unterthiner Thomas Unterthiner works in deep learning, on the basic building blocks of neural networks and on how architectures developed for one kind of data can be carried over to another. Their papers here ask how design choices inside a network affect training speed and generalization, and whether image recognition really depends on convolutional structure. Both questions are tested through experiments on standard image classification benchmarks. The most cited of these works is [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)](https://md.research.elamur.ai/free/en/papers/clevert-2015-fast-accurate-deep-network-learning/), from 2015. It introduced an activation function that passes positive inputs through unchanged and, for negative inputs, bends smoothly toward a fixed negative floor. Pulling average activations closer to zero gives some of the effect of batch normalization at lower cost, and the paper backs this with an analysis based on the unit natural gradient. The second is [An Image is Worth 16x16 Words](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/), from 2020. It fed images, cut into small square patches and embedded as a sequence, into a Transformer encoder borrowed almost unchanged from language processing, and called the result the Vision Transformer (ViT). The later work shifts attention from the behaviour of individual units to whole architectures and to training at scale. The ViT study reports that, when trained on ImageNet alone, the model falls slightly behind a ResNet of similar size. It attributes the gap to the missing assumptions of locality and translation that convolutions build in, and it examines what changes as training grows larger. Earlier bylines place Unterthiner at [Johannes Kepler University of Linz](https://md.research.elamur.ai/free/en/universities/johannes-kepler-university-of-linz/) from 2013 to 2019, then at [Google](https://md.research.elamur.ai/free/en/companies/google/) from 2020 to 2022. The most recent ones, from 2023 and 2024, are at [Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/). Co-authors include [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/), [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/), [Xiaohua Zhai](https://md.research.elamur.ai/free/en/authors/xiaohua-zhai/), Mostafa Dehghani and Jakob Uszkoreit. Thomas Unterthiner has 2 papers in our library, in 2015–2020; often with [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/), [Jakob Uszkoreit](https://md.research.elamur.ai/free/en/authors/jakob-uszkoreit/), [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/). 2 papers in our library · 2015–2020 [OpenAlex](https://openalex.org/A5083016682) · [ORCID](https://orcid.org/0000-0001-5361-3087) · [**h-index 29**](https://openalex.org/authors/A5083016682) · [56 works](https://openalex.org/works?filter=author.id:A5083016682) · [34,123 citations](https://openalex.org/works?filter=author.id:A5083016682&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [📄 2 works in our library](https://md.research.elamur.ai/free/en/authors/thomas-unterthiner/#papers) · [🔗 2,180 citations in our library](https://md.research.elamur.ai/free/en/authors/thomas-unterthiner/cited-by/) · [1 more in our library without a page yet](#also) · [⭐ Paper Star rank #613](https://md.research.elamur.ai/free/en/paper-stars/page/7/#star-613) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Thomas+Unterthiner) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Thomas+Unterthiner) · [dblp](https://dblp.org/search?q=Thomas+Unterthiner) · [ACL Anthology](https://aclanthology.org/search/?q=Thomas+Unterthiner) ## Affiliations [🎓**Johannes Kepler University of Linz** Linz · Austria 2013–2019 · 20 papers](https://md.research.elamur.ai/free/en/universities/johannes-kepler-university-of-linz/) · [🏢**Google** Mountain View · United States 2020–2022 · 8 papers](https://md.research.elamur.ai/free/en/companies/google/) · [🏢**Google DeepMind** London · United Kingdom 2023–2024 · 3 papers](https://md.research.elamur.ai/free/en/companies/google-deepmind/) · [🏢**RISC Software** Hagenberg im Mühlkreis · Austria 2016–2019 · 2 papers](https://ror.org/0533vxh11) · [🤝**Institute of Bioinformatics** Bengaluru · India 2016–2018 · 2 papers](https://ror.org/04hqfvm50) From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. ## Works by year ### 2020s [**2020** 1 paper](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/) ### 2010s [**2015** 1 paper](https://md.research.elamur.ai/free/en/papers/clevert-2015-fast-accurate-deep-network-learning/) ## Papers The most cited first (OpenAlex and our library). 1. **[Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)](https://md.research.elamur.ai/free/en/papers/clevert-2015-fast-accurate-deep-network-learning/)** *(2015 · 2,274 citations)* The paper proposes the exponential linear unit (ELU), an activation function for deep networks designed to train faster and generalize better than ReLU-style units. 2. **[An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/)** *(2020 · 1,947 citations)* The authors from Google Research test whether convolutional networks are truly necessary for image classification. ## Also in our library 1 more work of theirs is in our library without a page here yet; it will appear as we publish them. ## Co-authors here 5 co-authors with a page, the most papers together first. - [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/) - [Jakob Uszkoreit](https://md.research.elamur.ai/free/en/authors/jakob-uszkoreit/) - [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/) - [Mostafa Dehghani](https://md.research.elamur.ai/free/en/authors/mostafa-dehghani/) - [Xiaohua Zhai](https://md.research.elamur.ai/free/en/authors/xiaohua-zhai/) ## Cited by Papers of our library cite their works 2,180 times (of the 140,884 papers whose bibliographies we read); [90 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/thomas-unterthiner/cited-by/).