An Introduction to Convolutional Neural Networks

This paper is a brief tutorial introduction to Convolutional Neural Networks (CNNs), situating them within the recent rise of artificial neural networks (ANNs). It describes CNNs as biologically inspired computational models that have far exceeded earlier forms of artificial intelligence on common machine learning tasks, and explains that they are used primarily to solve difficult image-driven pattern recognition problems. The authors emphasize the CNN's precise yet relatively simple architecture as an accessible entry point into ANNs.

To orient newcomers, the document reviews recently published papers and newly formed techniques for developing CNN-based image recognition models, while assuming the reader already understands the fundamentals of ANNs and machine learning. As an accessible survey-style introduction rather than a new method, its value lies in lowering the barrier to entry and consolidating current practice for those beginning to work with convolutional networks.