Dimensionality reduction; F-transform; fuzzy partition; Laplacian eigenmaps; image restoration

The appropriate representation of high-dimensional data is the main focus of machine learning, pattern recognition and computer vision. With the same motivation, the F-transform uses fuzzy partitions in order to establish a compressed representation of data. Two distinguished properties of the F-transform: the best approximation in a local sense and dimensionality reduction contributed to the fact that the F-transform has many successful applications. We show that the technique of F-transform fully agrees with the technique of dimensionality reduction, based on Laplacian eigenmaps. To justify this claim, we characterize the processed by the F-transform data in terms of the adjacency graph that reflects their (data) intrinsic geometry. An application to the problem of image restoration is given.