Self-directed machine learning

Conventional machine learning (ML) relies heavily on manual design from experts to decide tasks, data, models, optimization algorithms, and evaluation metrics, which is labor-intensive, time-consuming, cannot learn autonomously like humans. In education science, self-directed learning, where human learners select tasks materials their own without requiring hands-on guidance, has been shown be more effective than passive teacher-guided learning. Inspired by the concept of we introduce principal Self-directed Machine Learning (SDML) propose a framework for SDML. Specifically, SDML as process guided self-awareness, including internal awareness external awareness. Our proposed benefits self task selection, data model strategy selection metric through self-awareness guidance. Meanwhile, performance serves feedback further improve self-awareness. We mathematical formulation based multi-level optimization. Furthermore, present case studies together with potential applications SDML, followed discussing future research directions. expect that could enable machines conduct human-like provide new perspective towards artificial general intelligence.