Extending greedy feature selection algorithms to multiple solutions

Abstract

Most feature selection methods identify only a single solution. This is acceptable for predictive purposes, but is not sufficient for knowledge discovery if multiple solutions exist. We propose a strategy to extend a class of greedy methods to efficiently identify multiple solutions, and show under which conditions it identifies all solutions. We also introduce a taxonomy of features that takes the existence of multiple solutions into account. Furthermore, we explore different definitions of statistical equivalence of solutions, as well as methods for testing equivalence. A novel algorithm for compactly representing and visualizing multiple solutions is also introduced. In experiments we show that (a) the proposed algorithm is significantly more computationally efficient than the TIE* algorithm, the only alternative approach with similar theoretical guarantees, while identifying similar solutions to it, and (b) that the identified solutions have similar predictive performance.

Liquid Biopsy in Type 2 Diabetes Mellitus Management: Building Specific Biosignatures via Machine Learning

“…To fully assess clinical relevance, we next plan to validate these biosignatures in independent, blinded validation cohorts of T2DM patients, obese and prediabetic individuals from a different clinical setting (cross-dataset analysis). We should note, however, that JADBio implements internal cross-validation with the bootstrap corrected cross-validation (BBC-CV) algorithm [36,37], shown to substitute external validation. The bootstrapping technique performs a correction to the estimation of out-of-sample performance of the final model.…”

“…In fact, JADBio performs this training and test procedure numerous times to reduce the variance of the estimation. Specifically, for small sample sizes, it employees a stratified, K-fold, repeated cross-validation protocol BBC-CV algorithm [37] that exhibits small estimation variance and removes the estimation bias due to the fact that it was selected among many. The features selected in the winning pipeline are the ones included in the returned model.…”

A Machine Learning Model to Predict Knee Osteoarthritis Cartilage Volume Changes over Time Using Baseline Bone Curvature

“…Second, in practice, for a given ML problem, multiple equivalent solutions in variable selections can exist [ 60 ]. A shortcoming of some variable selection methods is that they injudiciously identify only a single solution, minimizing a loss function like mean squared error, classification error, etc.…”

A Sparse Recovery Algorithm Based on Arithmetic Optimization

“…The position update of the operator in the AOA algorithm is carried out through its own experience and the neighborhood experience. In this paper, the algorithm AOA-CS is based on the population evolution strategy and process, and incorporates a random component to perform location updates [22][23][24][25][26]. Each time the operator starts searching, it first updates MOA and MOP values, and then generates three random numbers r1, r2 and r3 that obey uniform distribution between 0 and 1.…”