Exclusive: the most-cited papers of the twenty-first century (Category Science)

Interesting. article Nature, in which the authors tried to calculate the most cited articles 21 century. It was quite difficult to calculate accurately, since there is a whole bunch of resources that consider citations to be their own, but the authors took a balanced approach. In addition, in the field of computer science and ML, authors often publish preprints first, and then the same articles are published following the results of conferences, as a result, often references to the preprint version are not taken into account when calculating the citation of the main article.

It is interesting to note that in 25 The best articles got a lot of ML articles, here is their list with their places. 1. Deep Residual Learning for Image Recognition (ResNet), (2016), He, K., Zhang, X., Ren, S., & Sun, J. This article describes the ResNet architecture, which describes the architecture of neural networks with approximately 150 layers, approximately 5 more than was previously accepted. This architecture has overcome the problem of signal attenuation as the number of layers increases, which was a breakthrough for the development of deep learning and subsequent advances in AI such as AlphaGo, AlphaFold and ChatGPT. 6. Random forests ,(2001), Leo Breiman This work presents a machine-learning algorithm that has significantly improved previous methods of decision tree ensembles. This paper describes a new way to build an ensemble: many independent decision trees are trained on random subsamples of data and random subsets of features, and the final decision is made by a majority of votes. (classification) average (regression) 7. Attention Is All You Need, (2017), Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin In this paper, researchers from Google introduced the Transformer architecture, which revolutionized natural language processing and other sequential data. The main innovation was the rejection of recurrent neural networks in favor of the mechanism of self-attention. (self-awareness)It allows the model to determine the importance of each element of the sequence relative to others and to take into account the context at any distance. 8. ImageNet Classification with Deep Convolutional Neural Networks (2012), Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton This article described the architecture of AlexNet, a deep convolutional neural network. (CNN), who made a breakthrough in computer vision, winning the competition ILSVRC-2012 top-up5 into 15.3percentage 26.2The nearest competitor. AlexNet has demonstrated that deep CNNs trained on big data are able to recognize objects with high accuracy even without manually tweaking features, laying the foundations for modern deep learning. 12. U-Net: Convolutional Networks for Biomedical Image Segmentation (2015), Olaf Ronneberger, Philipp Fischer, Thomas Brox This article introduced a revolutionary architecture for biomedical image segmentation combining a coding pathway. (layering) and symmetrical decoding path (Transposed convolutions to restore resolution)connected by skip-links to save spatial information 16. Deep learning (2015), Yann LeCun, Yoshua Bengio, Geoffrey Hinton Review article about deep learning from the luminaries of this topic 24. ImageNet: A large-scale hierarchical image database (2009), Jia Deng; Wei Dong; Richard Socher; Li-Jia Li; Kai Li; Li Fei-Fei The article describes the creation of a large-scale image database ImageNet, organized by the hierarchy of WordNet. The purpose of the project is to collect most of the 80 000 WordNet synsets 500–1000 Full-length and carefully tagged images provide tens of millions of annotated examples. I wonder what in 8 AlexNet showed its top results while studying on ImageNet.

P.S. By the way, if you look at the rest of the top articles, there are many articles about medicine in general and cancer in particular, as well as about the software that was used for research.

#Science #ML #AI #Software #Engineering