This paper proposes new linear and nonlinear feature extractors based on the maximum margin criterion (MMC) for efficient and robust feature extraction in pattern recognition. Traditional methods like PCA and LDA have limitations, particularly in handling the small sample size problem. MMC aims to maximize the margin between classes after dimensionality reduction, leading to better class separability than PCA. By incorporating constraints, MMC can derive LDA and avoid the small sample size problem. The proposed methods are tested on various datasets, showing improved performance and stability compared to existing techniques. The linear feature extractor is derived by solving an eigenvalue problem, while the nonlinear version uses kernelization. Experimental results demonstrate that the new methods are effective, stable, and efficient, outperforming traditional approaches in tasks such as face recognition, cancer classification, and gene expression analysis. The paper also addresses the challenges of high-dimensional data and provides insights into the advantages of MMC over other methods.This paper proposes new linear and nonlinear feature extractors based on the maximum margin criterion (MMC) for efficient and robust feature extraction in pattern recognition. Traditional methods like PCA and LDA have limitations, particularly in handling the small sample size problem. MMC aims to maximize the margin between classes after dimensionality reduction, leading to better class separability than PCA. By incorporating constraints, MMC can derive LDA and avoid the small sample size problem. The proposed methods are tested on various datasets, showing improved performance and stability compared to existing techniques. The linear feature extractor is derived by solving an eigenvalue problem, while the nonlinear version uses kernelization. Experimental results demonstrate that the new methods are effective, stable, and efficient, outperforming traditional approaches in tasks such as face recognition, cancer classification, and gene expression analysis. The paper also addresses the challenges of high-dimensional data and provides insights into the advantages of MMC over other methods.