Software development for interactive visualization of high-dimensional data

Dimensionality reduction is the process of mapping high dimensional data to a lower dimensional space, such that some properties are preserved. The process of dimensionality reduction requires the scientist to take decisions and make compromises, as part of the information is typically lost in the process. This paper presents a software environment for dimensionality reduction. The program offers an extensive array of tools for quality assessment, as well as features aiming to develop the user's intuition and help the development of new dimensionality reduction algorithms. This work also introduces a novel approach to multidimensional scaling (the process of preserving the pairwise distances between the original space and lower dimensional space), which enables the user to make embeddings in O(N) time while consuming O(N) memory.