ÉtudeScience

Computer Science & AI · 2012

ImageNet Classification with Deep Convolutional Neural Networks

Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton

University of Toronto

Advances in Neural Information Processing Systems (NeurIPS), vol. 25 ·

Cited by 130,000+Open access
View original paper

Opens the version of record via DOI

'AlexNet' showed that deep convolutional neural networks, trained on GPUs, could crush the competition at large-scale image recognition. Its dramatic win on the ImageNet benchmark ignited the deep-learning revolution. (Linked DOI is the journal version of record; the original appeared at NeurIPS 2012.)

Sparked the modern deep-learning era in artificial intelligence.

  • A deep CNN nearly halved the error rate on the ImageNet image-classification benchmark.
  • Demonstrated GPU training, ReLU activations, and dropout regularisation at scale.
  • Proved deep learning could outperform hand-engineered computer-vision features.

A large convolutional neural network with five convolutional and three fully connected layers was trained on 1.2 million labelled ImageNet images using two GPUs, ReLU nonlinearities, dropout, and data augmentation.

Keywords

Computer Science & AI

Highly Accurate Protein Structure Prediction with AlphaFold

Jumper et al. · 2021 · Nature

AlphaFold solved a 50-year grand challenge in biology: predicting a protein's three-dimensional shape from its amino-acid sequence. Its accuracy approached that of experimental methods, transforming structural biology.

Cited by 37,000+Open access

Computer Science & AI

Attention Is All You Need

Vaswani et al. · 2017 · Advances in Neural Information Processing Systems (NeurIPS)

The paper that introduced the Transformer, an architecture built entirely on attention mechanisms with no recurrence or convolution. It became the foundation of modern large language models such as GPT and BERT.

Cited by 180,000+Open access

Computer Science & AI

Deep Learning

LeCun, Bengio & Hinton · 2015 · Nature

A landmark review by three pioneers of the field, synthesising how deep neural networks learn layered representations of data. It explains why depth works and surveys the advances that made deep learning dominant across vision, speech, and language.

Cited by 110,000+

Étude Science indexes and summarises this work; it is not the publisher. The summary above is written by Étude. For the definitive text, figures, and data, please consult the original publication via the link above. Krizhevsky, Sutskever & Hinton (2012) hold the rights to the original work.