Model reduction and training efficiency in computer-based reasoning and artificial intelligence systems

This invention describes a way to make artificial intelligence models more efficient and balanced by carefully selecting the data they use. It works by calculating "conviction scores" for each piece of data, which measure its relevance or familiarity, and then decides whether to include it in the AI model based on specific conditions and thresholds. This process helps create leaner, more effective AI models for controlling various systems, from self-driving cars to smart voice assistants. The claims specifically detail using "distance contribution" and "familiarity conviction" scores.