Scalable Nonlinear Learning with Adaptive Polynomial Expansions
Alekh Agarwal, Alina Beygelzimer, Daniel Hsu, John C. Langford, Matus Telgarsky · arXiv (Cornell University) · 2014
Can we effectively learn a nonlinear representation in time comparable to linear learning? We describe a new algorithm that explicitly and adaptively expands higher-order interaction features over base linear representations. The algorithm is designed for extreme computational efficiency, and an extensive experimental study shows that its computation/prediction tradeoff ability compares very favorably against strong baselines.