What is PRISM™
PRISM™ (Predictive Risk Identification and Systemic Mitigation) is Albeado's proprietary AI/ML framework which delivers "data driven yet explainable" analysis and predictions for complex interconnected systems.
PRISM™ has been designed to understand and explain the gaps between real-time observation of an actual operational system, and designers' and planners' understanding of how the system should behave at any given moment.
PRISM™ is built atop two unique award-winning technology stacks, namely,
- DSCube™ (Deep State Space Synthesis), which lets planners, operators, designers and support engineers determine from observed data what a large complex system - like a 5G/LTE mobile network, or even evolution of cancer in the human body - is doing at any given moment of time. And even more importantly, what it will do in the future.
- XNML™ (eXplainable Network based Machine Learning) is an underlying system discovery and data driven modeling system which reveals, through computational graph topology and metrics, critical relationships and strength of influences of interconnected events and actions.
How does PRISM™ "explain" business operations?
Explainability is generated in terms of state and transition descriptors that measure and annotate meaningful metrics and their gradients over all the system states observed in the data and cohorts. Pathways potentially leading to states in a Region-Of-Interest (ROI) such as malfunctioning or lock down states in LTE networks, provide path level motif from any observed or latent state in the DSCube™ purview all the way to a potential failure or malfunctioning state.
Labels formed from such state and transition descriptors are then composed in different cohorts, taxonomies and hierarchies to generate higher level systemic descriptors. Such new (super)label generation, which are really elements of higher level of explanations, thus iteratively enriches and refines the emergent explanations.
PRISM™ also predicts the likelihood of encountering a degraded or a failure state as well as the expected time to failure (as part of advanced alert generation) based on historical event modeling.
Can PRISM™ be used for Root Cause Analysis (RCA)?
Absolutely. The DSCube™ product, with help from XNML™ based system discovery, inherently links observed states to failure states. State transition trajectories observed in XNML™ thereby quantifies pathways that lead to a malfunction (as stated above). With the human-in-the-loop paradigm, annotations of observed state as well as historical event analysis helps create a framework for attributing root cause for observed malfunction/failure. DSCube™ is designed to do attributions both at metric level (based on performance indicators) as well as function and module level.
How unique is PRISM™?
PRISM™ has undergone deep due diligence by scientific (NSF, DOE), industrial (CSI, IEEE) and business communities and adjudged to be unique in its segment of "data driven yet explainable AI/ML for complex systems".
That has brought Albeado multi-million dollars of competitive awards and grants to accelerate development of products and solutions around it as well as prestigious business engagements.
Primarily, our methodology differentiates itself by focusing on system state space quantification as opposed to data feature space modeling and criticality attribution.