Feature selection using stochastic approximation with Barzilai and Borwein non-monotone gains
With recent emergence of machine learning problems with massive number of features, feature selec... more With recent emergence of machine learning problems with massive number of features, feature selection (FS) has become an ever-increasingly important tool to mitigate the effects of the so-called curse of dimensionality. FS aims to eliminate redundant and irrelevant features for models that are faster to train, easier to understand, and less prone to overfitting. This study presents a wrapper FS method based on Simultaneous Perturbation Stochastic Approximation (SPSA) with Barzilai and Borwein (BB) non-monotone gains within a pseudo-gradient descent framework wherein performance is measured via cross-validation. We illustrate that SPSA with BB gains (SPSA-BB) provides dramatic improvements in terms of the number of iterations for convergence with minimal degradation in cross-validated error performance over the current state-of-the art approach with monotone gains (SPSA-MON). In addition, SPSA-BB requires only one internal parameter and therefore it eliminates the need for careful fi...
This manuscript presents the following: (1) an improved version of the Binary Simultaneous Pertur... more This manuscript presents the following: (1) an improved version of the Binary Simultaneous Perturbation Stochastic Approximation (SPSA) Method for feature selection in machine learning (Aksakalli and Malekipirbazari, Pattern Recognition Letters, Vol. 75, 2016) based on non-monotone iteration gains computed via the Barzilai and Borwein (BB) method, (2) its adaptation for feature ranking, and (3) comparison against popular methods on public benchmark datasets. The improved method, which we call SPSA-FSR, dramatically reduces the number of iterations required for convergence without impacting solution quality. SPSA-FSR can be used for feature ranking and feature selection both for classification and regression problems. After a review of the current state-of-the-art, we discuss our improvements in detail and present three sets of computational experiments: (1) comparison of SPSA-FS as a (wrapper) feature selection method against sequential methods as well as genetic algorithms, (2) com...
Earthquake relief network involves storage and distribution of relief aid to people in need. In t... more Earthquake relief network involves storage and distribution of relief aid to people in need. In this paper, a new stochastic multi-objective mixed integer mathematical model is developed and implemented in Kadikoy municipality of Istanbul, Turkey in order to configure part of the earthquake relief network. The aim of the model is to help decision makers decide on the locations of storage areas for shelters pre-earthquake and distribution of shelters from these areas to temporary shelter areas post-earthquake while minimizing earthquake scenario-specific total expected distribution distance, total expected earthquake damage risk factor of storage areas and expected total penalty cost related to unsatisfied demand at temporary shelter areas, simultaneously. In the model, storage area capacity and coverage distance restrictions are taken into consideration. The data related to potential storage areas and shelter locations were obtained from Kadikoy municipality of Istanbul and Istanbul...
Earthquake relief network involves storage and distribution of relief aid to people in need. In t... more Earthquake relief network involves storage and distribution of relief aid to people in need. In this paper, a new stochastic multiobjective mixed integer mathematical model is developed and implemented in Kadikoy municipality of Istanbul, Turkey in order to con gure part of the earthquake relief network. e aim of the model is to help decision makers decide on the locations of storage areas for shelters pre-earthquake and distribution of shelters from these areas to temporary shelter areas post-earthquake while minimizing earthquake scenario-speci c total expected distribution distance, total expected earthquake damage risk factor of storage areas and expected total penalty cost related to unsatis ed demand at temporary shelter areas, simultaneously. In the model, storage area capacity and coverage distance restrictions are taken into consideration. e data related to potential storage areas and shelter locations were obtained from Kadikoy municipality of Istanbul and Istanbul Metropolitan Municipality (IMM). e earthquake damage risk factors were determined based on possible earthquake scenarios given in Japan International Cooperation Agency's (JICA) report. Four event scenarios with two di erent earthquake scenario likelihoods were considered and sample e cient solutions from the Pareto frontier were obtained implementing the normalized (scaled) weighted sum method.
Uploads
Papers by Zeren Yenice