Presentation: Multi-Layered Vector Spaces

Analyzing and classifying arbitrary sequences based on structural similarities and differences, no matter how subtle, is a mathematical problem of escalating relevance and surging importance in many scientific disciplines. One of the many challenges in designing classification algorithms for arbitrary sequences is the extraction and representation of significant features from the data. On potential solution to the characterization of sequence data is a simple mathematical model, referred to as Multi-Layered vector Spaces (MLVS). This model coverts a sequence into a large number of vectors that preserves location and distribution information between arbitrary pairs of elements; the vectors, in turn, are easily exploitable by classification techniques. This talk will discuss the model and ongoing experiments within the biological domain to determine the utility of this representation.