LLNL has expertise in both applying and extending a wide variety of state-of-the-art Machine Learning algorithms, including Neural Networks, Random Forests, and Dynamic Belief Networks. Our researchers have used these algorithms to tackle complex classification, clustering, change and anomaly detection problems on diverse data types including network sensor information, natural language, and imagery/video. With the experience and infrastructure to handle “big data” (batch and streaming), we leverage large, highly-complex and multi-modal datasets to produce actionable information.
Probabilistic topic modeling such as Latent Dirichlet Allocation yields valuable insight into anomalous traffic patterns, providing cues for analysts.