ImageNet Classification with Deep Convolutional Neural Networks

By Alex Krizhevsky et al

Read the original document by opening this link in a new tab.

Table of Contents

1 Introduction
2 The Dataset
3 The Architecture
3.1 ReLU Nonlinearity
3.2 Training on Multiple GPUs
3.3 Local Response Normalization
3.4 Overlapping Pooling
3.5 Overall Architecture
4 Reducing Overfitting
4.1 Data Augmentation

Summary

This paper presents the training of a large, deep convolutional neural network on the ImageNet dataset for object recognition. The network achieved top-1 and top-5 error rates significantly better than previous methods. The architecture includes novel features like ReLU nonlinearity, training on multiple GPUs, local response normalization, and overlapping pooling. Data augmentation techniques such as image translations and intensity alterations were used to combat overfitting. The paper discusses the importance of large datasets for training models and the overall structure of the network.