← Back to papers
2014 NIPS NeurIPS 2014

Scalable Non-linear Learning with Adaptive Polynomial Expansions

Abstract

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.

🧭 Keyword Pioneer — nonlinear representation
🐝 Cross-Pollinator — Artificial Intelligence, Computer Science, Computer Vision, Deep Learning, Machine Learning, Mathematics & Optimization, Natural Language Processing, Reinforcement Learning, Speech & Audio
🌉 Interdisciplinary Bridge — Artificial Intelligence and Machine Learning
📈 Trend Setter — Efficient Computing
🐣 Hot Topic Early Bird — computational efficiency

Authors

Alekh Agarwal , Alina Beygelzimer , Daniel J. Hsu , John Langford , Matus J. Telgarsky , Matus J Telgarsky

Topics

Machine Learning > Core Methods > Representation Learning Machine Learning > Optimization & Theory > Optimization Machine Learning > Learning Types > Representation Learning Machine Learning > Core Methods > Feature Learning Artificial Intelligence > Core AI > Efficient Computing Machine Learning > Core Methods > Optimization Machine Learning > Learning Types > Feature Learning

Keywords

representation learning adaptive learning computational efficiency feature expansion adaptive algorithm nonlinear representation polynomial expansions feature engineering polynomial feature nonlinear learning non-linear learning
Download PDF

Related papers

Information-based learning by agents in unbounded state spaces 2014
Stochastic Gradient Descent, Weighted Sampling, and the Randomized Kaczmarz algorithm 2014
Partition-wise Linear Models 2014
Active Regression by Stratification 2014
Cone-Constrained Principal Component Analysis 2014