Fast Training of Convolutional Networks through FFTs
This addresses the problem of high computational costs for training and inference in convolutional networks, particularly for large datasets in computer vision and machine learning, representing a strong specific gain rather than a foundational change.
The paper tackles the slow training and inference of convolutional networks by proposing an algorithm that accelerates these processes through FFTs, achieving improvements of over an order of magnitude compared to state-of-the-art implementations.
Convolutional networks are one of the most widely employed architectures in computer vision and machine learning. In order to leverage their ability to learn complex functions, large amounts of data are required for training. Training a large convolutional network to produce state-of-the-art results can take weeks, even when using modern GPUs. Producing labels using a trained network can also be costly when dealing with web-scale datasets. In this work, we present a simple algorithm which accelerates training and inference by a significant factor, and can yield improvements of over an order of magnitude compared to existing state-of-the-art implementations. This is done by computing convolutions as pointwise products in the Fourier domain while reusing the same transformed feature map many times. The algorithm is implemented on a GPU architecture and addresses a number of related challenges.