Scalable Nonlinear Learning with Adaptive Polynomial Expansions

Dec 8, 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.

  • Neural Information Processing Systems (NIPS)
  • Conference/Workshop Paper