Subsampling is the reduction of data resolution or volume by selecting a subset or summary of the original inputs. In machine learning contexts, it is often associated with pooling or sampling operations that reduce dimensionality while retaining useful information.