A General Purpose Toolkit for Exploiting Network Data in Diverse Applications

Stottler Henke proposes to develop a generalized toolset around a new and innovative network data encoding strategy with the goals of: (1) increasing the scalability of network analysis and visualization tools; (2) fundamentally improving the intelligibility of network models and thereby increasing the accessibility of network science to the broader research community; and (3) providing a more faithful representation of reality in situations where generic modeling approaches may obscure or misrepresent system dynamics and engineered functions. The proposed system, called NetSet, will be comprised of three primary components. First, it will provide the capability to automatically identify recurring and meaningful structures (e.g., Motifs) within complex networks in order to improve the scalability, accuracy, and exploitability of network models. Second, NetSet will move beyond the typical unimodal (i.e., nodes and links) representation and capture broader descriptive features. Last, the proposed system will bridge the current gap between traditional network science and Statistical Relational Learning so as to support a broader range of predictive analysis applications. Our Phase I prototyping will prove the feasibility of our approach and set the stage for the complete implementation of NetSet in Phase II and its subsequent commercialization.