truncation SVD is a popular method for dimensionality reduction. However, it works better with sparse data. A given m⤫n matrix, truncated SVD will produce matrices with the specified number of columns.
