Zobrazeno 1 - 10
of 111
pro vyhledávání: '"Collier, Mark P."'
Accurate uncertainty estimation is vital to trustworthy machine learning, yet uncertainties typically have to be learned for each task anew. This work introduces the first pretrained uncertainty modules for vision models. Similar to standard pretrain
Externí odkaz:
http://arxiv.org/abs/2402.16569
Autor:
Fadeeva, Anastasiia, Schlattner, Philippe, Maksai, Andrii, Collier, Mark, Kokiopoulou, Efi, Berent, Jesse, Musat, Claudiu
The adoption of tablets with touchscreens and styluses is increasing, and a key feature is converting handwriting to text, enabling search, indexing, and AI assistance. Meanwhile, vision-language models (VLMs) are now the go-to solution for image und
Externí odkaz:
http://arxiv.org/abs/2402.15307
Autor:
Wang, Ke, Ortiz-Jimenez, Guillermo, Jenatton, Rodolphe, Collier, Mark, Kokiopoulou, Efi, Frossard, Pascal
Label noise is a pervasive problem in deep learning that often compromises the generalization performance of trained models. Recently, leveraging privileged information (PI) -- information available only during training but not at test time -- has em
Externí odkaz:
http://arxiv.org/abs/2310.06600
Autor:
Kossen, Jannik, Collier, Mark, Mustafa, Basil, Wang, Xiao, Zhai, Xiaohua, Beyer, Lucas, Steiner, Andreas, Berent, Jesse, Jenatton, Rodolphe, Kokiopoulou, Efi
We introduce Three Towers (3T), a flexible method to improve the contrastive learning of vision-language models by incorporating pretrained image classifiers. While contrastive models are usually trained from scratch, LiT (Zhai et al., 2022) has rece
Externí odkaz:
http://arxiv.org/abs/2305.16999
Autor:
Ortiz-Jimenez, Guillermo, Collier, Mark, Nawalgaria, Anant, D'Amour, Alexander, Berent, Jesse, Jenatton, Rodolphe, Kokiopoulou, Effrosyni
Leveraging privileged information (PI), or features available during training but not at test time, has recently been shown to be an effective method for addressing label noise. However, the reasons for its effectiveness are not well understood. In t
Externí odkaz:
http://arxiv.org/abs/2303.01806
Autor:
Dehghani, Mostafa, Djolonga, Josip, Mustafa, Basil, Padlewski, Piotr, Heek, Jonathan, Gilmer, Justin, Steiner, Andreas, Caron, Mathilde, Geirhos, Robert, Alabdulmohsin, Ibrahim, Jenatton, Rodolphe, Beyer, Lucas, Tschannen, Michael, Arnab, Anurag, Wang, Xiao, Riquelme, Carlos, Minderer, Matthias, Puigcerver, Joan, Evci, Utku, Kumar, Manoj, van Steenkiste, Sjoerd, Elsayed, Gamaleldin F., Mahendran, Aravindh, Yu, Fisher, Oliver, Avital, Huot, Fantine, Bastings, Jasmijn, Collier, Mark Patrick, Gritsenko, Alexey, Birodkar, Vighnesh, Vasconcelos, Cristina, Tay, Yi, Mensink, Thomas, Kolesnikov, Alexander, Pavetić, Filip, Tran, Dustin, Kipf, Thomas, Lučić, Mario, Zhai, Xiaohua, Keysers, Daniel, Harmsen, Jeremiah, Houlsby, Neil
The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image an
Externí odkaz:
http://arxiv.org/abs/2302.05442
Autor:
Collier, Mark, Jenatton, Rodolphe, Mustafa, Basil, Houlsby, Neil, Berent, Jesse, Kokiopoulou, Effrosyni
Heteroscedastic classifiers, which learn a multivariate Gaussian distribution over prediction logits, have been shown to perform well on image classification problems with hundreds to thousands of classes. However, compared to standard classifiers, t
Externí odkaz:
http://arxiv.org/abs/2301.12860
Autor:
Tran, Dustin, Liu, Jeremiah, Dusenberry, Michael W., Phan, Du, Collier, Mark, Ren, Jie, Han, Kehang, Wang, Zi, Mariet, Zelda, Hu, Huiyi, Band, Neil, Rudner, Tim G. J., Singhal, Karan, Nado, Zachary, van Amersfoort, Joost, Kirsch, Andreas, Jenatton, Rodolphe, Thain, Nithum, Yuan, Honglin, Buchanan, Kelly, Murphy, Kevin, Sculley, D., Gal, Yarin, Ghahramani, Zoubin, Snoek, Jasper, Lakshminarayanan, Balaji
A recent trend in artificial intelligence is the use of pretrained models for language and vision tasks, which have achieved extraordinary performance but also puzzling failures. Probing these models' abilities in diverse ways is therefore critical t
Externí odkaz:
http://arxiv.org/abs/2207.07411
Supervised learning datasets often have privileged information, in the form of features which are available at training time but are not available at test time e.g. the ID of the annotator that provided the label. We argue that privileged information
Externí odkaz:
http://arxiv.org/abs/2202.09244
Autor:
Fortuin, Vincent, Collier, Mark, Wenzel, Florian, Allingham, James, Liu, Jeremiah, Tran, Dustin, Lakshminarayanan, Balaji, Berent, Jesse, Jenatton, Rodolphe, Kokiopoulou, Effrosyni
Uncertainty estimation in deep learning has recently emerged as a crucial area of interest to advance reliability and robustness in safety-critical applications. While there have been many proposed methods that either focus on distance-aware model un
Externí odkaz:
http://arxiv.org/abs/2110.02609