Welcome to the Home of SPSA-FSR: SPSA for Feature Selection and Ranking (FSR) in Machine Learning
SPSA (Simultaneous Perturbation Stochastic Approximation)-FSR is a competitive new method for feature selection and ranking in machine learning. It is an extension of a general-purpose black-box stochastic optimization algorithm, SPSA, applied to the FSR problem.
SPSA-FSR has been shown to outperform most of the state-of-the-art FSR methods (including ReliefF, mRMR, and Random Forest Importance) on challenging supervised machine learning problems, sometimes outperforming even the feature extraction method PCA (Principal Components Analysis). For further information on SPSA-FSR, please feel free to have a look at the following links:
For a quick overview and comparison of SPSA-FSR applied to feature ranking, please visit our tutorial here.
This site also contains comprehensive tutorials on (1) the Python programming language for data analytics, (2) introductory statistics, and (3) machine learning: