Is this you?
Sign in with ORCID to claim this profile — instantly verify it’s yours, curate your publications, and take control of how your work appears here.
Recent publications · auto-generated until this profile is claimed
Geoffrey E. Hinton’s indexed work has fundamentally reshaped machine learning, beginning with his first-author 2006 *Science* paper on reducing dimensionality with neural networks and a companion *Neural Computation* paper introducing a fast learning algorithm for deep belief nets, which together revived interest in deep architectures. As senior author, he led the 2015 *Nature* review "Deep learning" (over 82,000 citations) and the landmark 2017 *Communications of the ACM* paper on ImageNet classification with deep convolutional neural networks (over 75,000 citations), which demonstrated the power of large-scale supervised learning. His 2008 *Journal of Machine Learning Research* paper on t-SNE for visualizing high-dimensional data (over 35,000 citations) remains a standard tool for data exploration. Beyond these lead-author contributions, Hinton’s 2014 paper "Dropout: a simple way to prevent neural networks from overfitting" (over 34,000 citations) introduced a widely adopted regularization technique. His other first-author works include distilling knowledge in neural networks (2015, over 14,000 citations), rectified linear units for restricted Boltzmann machines (2010, over 13,000 citations), and improving neural networks by preventing co-adaptation of feature detectors (2012, over 6,600 citations). As senior author, he contributed to a simple framework for contrastive learning of visual representations (2020, over 7,300 citations). Across these works, Hinton’s research arc has consistently advanced both the theoretical foundations and practical capabilities of deep learning, from unsupervised pretraining to modern representation learning.
0 publications
0 other materials
No GitHub profile linked
Alpha1 Science activity will appear here once this profile is claimed.