Applied R&D, Machine Learning & Mathematical Modeling
Architecting foundational mathematical frameworks, graph algorithms, and scalable statistical models for high-dimensional, noisy data systems.
Core Technical Research Pillars
1. Graph AI, Network Science & Dynamic Clustering
Developing robust algorithms for community detection, constrained graph partitioning, and signal tracking in large-scale networks subject to measurement noise.
- Constrained community detection in social networks. W. D. Viles and A. J. O’Malley. The New England Journal of Statistics in Data Science, 2(3):368–379, 2024.
[DOI]- Technical Impact: Formulated constrained optimization algorithms for graph partitioning under topological and node-level operational boundaries.
- Robust dynamic community detection with applications to human brain functional networks. L.-E. Martinet, M. A. Kramer, W. D. Viles, et al. Nature Communications, 11:2785, 2020.
[DOI]- Technical Impact: Engineered noise-resilient clustering algorithms to track temporal state transitions and dynamic network evolution in high-dimensional time-series graphs.
- On the propagation of low-rate measurement error to subgraph counts in large networks. P. Balachandran, E. D. Kolaczyk, and W. D. Viles. Journal of Machine Learning Research, 18(61): 1–33, 2017.
[Paper]- Technical Impact: Quantified mathematical error bounds for graph motif analysis, establishing reliable network analytics in noisy, real-world data pipelines.
2. Phase Transitions, Statistical Physics & Network Dynamics
Designing mathematical models to infer critical phase transitions and percolation dynamics in noisy, evolving complex systems.
- Percolation under noise: Detecting explosive percolation using the second-largest component. W. Viles, C. E. Ginestet, A. Tang, M. A. Kramer, and E. D. Kolaczyk. Physical Review E, 93(5): 052301, 2016.
[DOI]- Technical Impact: Formulated statistical inference frameworks and hypothesis tests to detect explosive percolation phase transitions in noisy, dynamically evolving graph structures.
3. Complex Systems & High-Dimensional Biological Analytics
Modeling high-order interactions and multi-modal feature spaces in sparse, irregularly sampled observational data.
- Identifying stationary microbial interaction networks based on irregularly spaced longitudinal 16S rRNA gene sequencing data. J. Zhou, J. Gui, W. D. Viles, et al. Frontiers in Microbiomes, 3, 2024.
[DOI]- Technical Impact: Constructed longitudinal network inference frameworks tailored for asynchronous sampling intervals and non-stationary dynamic systems.
- Information content of high-order associations of the human gut microbiota network. W. D. Viles, J. C. Madan, H. Li, M. R. Karagas, and A. G. Hoen. The Annals of Applied Statistics, 15(4):1788–1807, 2021.
[JSTOR]- Technical Impact: Designed higher-order graphical model estimators capable of capturing non-pairwise, hyper-graph dependency structures in multivariate counts.
Funded R&D & Cross-Functional Research Leadership
Co-Investigator | Multi-omic Functional Integration using Networks
National Institutes of Health (NIH) / National Library of Medicine | 2017 – 2021
- Project Scope: Led quantitative methodology development for integrating multi-omic biological data streams using graph-based computational pipelines.
- Deliverables: Built scalable network inference algorithms to isolate key statistical associations in complex biological ecosystems, facilitating automated hypothesis generation for clinical domain experts.
Technical Capability Summary Matrix
| Domain Pillar | Core Methodologies | Industry Value & Application |
|---|---|---|
| Graph & Network Analytics | Community Detection, Motif Analysis, Dynamic Graphs, Error Bounds | Social Network Analysis, Fraud/Entity Resolution, Brain-Computer Interfaces |
| Network Dynamics & Physics | Phase Transitions, Percolation Modeling, Noisy Systems, Graph Phase Inferences | Anomaly Detection, System Stability Modeling, Reliability Engineering |
| Complex Systems | Hyper-graph Associations, Longitudinal Inference, High-Dimensional Data | Bioinformatics, Health Tech, Sensor Stream Analytics, Time-Series Forecasting |