ImageNet Classification with Deep Convolutional Neural Networks

This paper presents AlexNet, a deep convolutional neural network architecture that achieved a top-5 error rate of 15.3% on the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) 2012, significantly outperforming traditional computer vision methods. The network consists of five convolutional layers followed by three fully connected layers, trained on two GPUs using techniques like ReLU activation, dropout regularization, and data augmentation.

The authors—Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton—demonstrated that deep learning could dramatically improve image classification performance on large-scale datasets. The paper's success sparked widespread adoption of convolutional neural networks across computer vision and beyond, establishing deep learning as the dominant paradigm in artificial intelligence.

The work's impact extends far beyond its immediate results: it validated the potential of neural networks at scale, showed the importance of GPU computing for training large models, and opened a new era of AI research. AlexNet is widely credited as the catalyst for the deep learning revolution of the 2010s.