Convolution networks

Convolution networks, also called convolutional neural networks (CNNs), are like neural networks for processing data in the form of a grid. This grid can be 2D, such as an image, 1D, such as a time series, 3D, such as data from LIDAR sensors, and so on. The basic idea behind CNNs is inspired by how human vision works. In 1979, Fukushima proposed Neocognitron. It was a one-of-a-kind architecture that was directly inspired by how human vision works. But CNNs came into existence as we know them today in 1989 when Yann LeCun used backpropagation to learn such a network and proved it by getting state-of-the-art results in handwritten digit recognition. In 2012, when AlexNet (a CNN architecture for image recognition) won the annual challenge of image recognition called ImageNet, that too by a large margin between it and competing non-deep learning approaches, the interest and research in CNNs peaked. People soon figured out that, apart from images, CNNs are effective...