Output details

Density-Preserving Sampling: Robust and Efficient Alternative to Cross-Validation for Error Estimation

The proposed approach to estimating predictive model generalisation reduces required computations by an order of magnitude when compared to the industry standard. Implementation of the method has been included in PRTools (Pattern Recognition Toolbox) for MATLAB developed at TU Delft with 500+ downloads a month. This impacts machine learning researchers and practitioners worldwide by enabling them to perform more experiments in the same amount of time, leading to more accurate predictive models. The method was also used in designing the winning solution of the ISMIS 2011 Music Information Retrieval competition, resulting in an invited presentation at the conference.