On the application of GP to streaming data classification tasks with label budgets

摘要: A framework is introduced for applying GP to streaming data classification tasks under label budgets. This a fundamental requirement if going adapt the challenge of environments. The proposes three elements: sampling policy, subset and archiving policy. policy establishes on what basis sampled from stream, therefore when information requested. used define individuals evolve against. composition such mixture forwarded historical identified through combination achieve decoupling between rate at which stream passes evolution commences. Benchmarking performed two artificial sets with specific forms sudden shift gradual drift as well known real-world set.