Tag
This paper analyzes why machine learning, particularly neural networks, remains opaque in its learning process by framing it as a complex dynamical system, identifying three key properties that contribute to learning opacity, and arguing that some sources may be irreducible.