Earlier work includes synthetic visualization of massive time-series. . Applications include near real-time anomaly detection, forensics data analysis, and support of machine learning through labelling of examples.
This work visualizes massive data through data selection, transformation and rendering. The visualization is one or more three dimensional unit volumes. This world model facilitates measurement and comparison between systems over multiple operating conditions. Intrinsic schema-methods recover structure from multiple data formats. Tools include a GUI-based user console useful in example-based filtration, with applications to forensics as well as the labelling of training instances for machine learning. This supports console-based usage and creation of streaming video clips. The video clips at this web site show a few small examples of the dynamic traffic and the benefits of streaming visualization. A separate image-based pattern recognizer measures image similarity, for clustering and anomaly detection. Some of this work began under partial support from AT&T Laboratories. |
