Differentiating between classes of objects is an important problem in content based image retrieval. To aid in this task, we are using dictionary learning to learn useful features (highlighted in red boxes) across categories of images that will lead to higher classification rates.
Our staff enjoy ample opportunity to publish technical work and present at conferences around the world. PNNL's unique cross-disciplinary perspective provides
practical solutions to critical challenges for both industry and government clients.
As leaders in applied statistics and mathematics research, our unique perspective provides practical solutions to important problems. Our cross-cutting capabilities among mathematical and statistical domains creates powerful and dynamic teams working to solve some of our nation's critical challenges using:
- Data and feature extraction
- Stochastic/deterministic scientific modeling
- Simulation and uncertainty management
- Analytics of non-traditional data
- Systematic planning and sampling design
- Experimental design and engineering statistics
- Anomaly detection and forecasting
- Decision science and operations research
The Risk Reduction and Resource Assessment Model (3RAM) assesses risks associated with terrorist threats to a transportation system. The model provides a way to gauge the relative risk and associated risk reduction as a result of alternative security strategies and serves as a decision support tool by automatically providing recommended resource allocation based on system inputs variables as a function of time.
PNNL has developed Visual Sample Plan (VSP), a software tool for determining the optimal number and placement of samples and for statistical analysis
of sample results to support confident decisions. With over 5000 users worldwide, VSP is becoming the premier sampling design and analysis software tool for environmental applications.