Veritone CDI

avtor: Veritone, Inc.

Advanced AI enabling near-optimal and near real-time decision making on large-scale data sets

The Cooperative Distributed Inferencing (CDI) system is a unique advanced technology enabling near-optimal and near real-time decision making on large-scale, heterogeneous and distributed information with the use of rules. CDI integrates absolute, hard and soft rules within complex knowledge-based decision support systems to achieve performance goals while satisfying various requirements from natural or governing laws, policies, and best practices.

The CDI system has a Distributed Architecture (DA), consisting of a network of Decision Elements (DEs) that work together to resolve queries and identify Pareto efficient states. The decision elements access shared information from both an Internal Heterogeneous Database (IHDB) and External Knowledge Base (EKB). Each decision element solves a query using optimal control theory, starting with a technique called analytic continualization - transforming the query and rules into differential equations whose dependent variables represent internal variables and parameters of the rules. The decision elements in the architecture are synchronized via a Pareto multi-criteria optimization strategy

CDI features a self-adapting and learning design. Since CDI converts the original query into an optimal control problem, it can use feedback from the environment (e.g., external sensors or internal knowledge updates from other DEs) to refine its internal model; the Hamilton-Jacobi-Bellman equation will be updated to reflect new information and automatically form soft rule-like constraints internally.

CDI is particularly applicable when the system has large-scale heterogeneous data, rules from government compliances and/or business requirements, and the need to make near real-time decisions. Healthcare and energy are two such applications.

Hiter pregled