What does stride refer to in a Convolutional Neural Network (CNN)?

In the context of Convolutional Neural Networks (CNNs), the term "stride" specifically refers to the number of steps taken before applying the convolutional filter again. When a filter is applied to an input image, the stride determines how much the filter moves across the image after each application. For example, if the stride is set to 1, the filter moves one pixel at a time across the width and height of the input. Conversely, if the stride is set to 2, the filter moves two pixels at a time. This movement affects the size of the output feature map generated by the convolutional layer.

Choosing an optimal stride is essential, as it influences both the computational efficiency of the CNN and how much spatial information is retained in the feature maps. A larger stride results in a smaller output size, while a smaller stride preserves more spatial information but requires more computation.

Understanding stride is crucial for anyone working with CNNs, as it directly impacts the architecture and performance of convolutional networks in tasks such as image recognition or processing.