Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Distributed and Online maintenance of graphical models in Apache Flink

Tzimos Nikolaos

Full record

---

Summary

With the growing need for large scale data analysis, distributed machine learning has grown importance in recent years. The raw data is described by large number of interrelated variables and an important task is to describe the joint probability distribution over these variables, allowing simultaneously interferences and predications to be made. Directly modeling of joint probability distribution of all these variables may be infeasible, since the complexity of such model grown exponential with the number of variables. We focus on Bayesian Networks, the father of graphical models and present a different communication-efficient approach using the well-known method of Functional Geometric Monitoring, for continuously learning and maintenance of Bayesian Networks in a distributed streaming environment. Finally, the experimental results confirmed the functionality of proposed method.

Available Files

Services