Cover’s theorem

Cover's theorem states that a complex pattern-classification problem cast in a high-dimensional space is more likely to be linearly separable than in a low-dimensional space.