Technical Topic 3.2.2.d Bayesian and Non-Parametric Statistics: Integration of Neural Networks with Bayesian Networks for Data Fusion and Predictive Modeling
Technical Report,15 Apr 2014,14 Jan 2015
West Virginia University Research Corporation Morgantown United States
Pagination or Media Count:
This was a short-term proof-of-concept project with the goal of demonstrating the feasibility of, and lay the theoretical foundations for, integration of predictive neural networks into Bayesian networks as a means of generating probability distribution functions and likelihood tables. The challenges were two-fold first, developing a way to convert XY data output from an instrument to a probability density functionusing a neural network and secondly, fusing this and other types of sensor output into a single probabilistic evaluation of multiple sensor outputs. Ultimately, this would be useful in application such as networked sensor arrays such as might be deployed to detect chemical agentsin a subway system for example.
- Statistics and Probability