Gradient Descent Batch Clustering for Image Classification

The batch clustering algorithm for classification application requires the initial parameters and also has a drifting phenomenon stochastic process. are critical to con-verge partial optimum. in original still space be improved thus speed up convergence based on parameters. This paper proposes an unsupervised method by addressing these two issues. Firstly, estimation been given preliminary with hierarchical manner of principal component analysis (PCA). nonlinear have estimated mathematical connection between PCA clusters membership. With parameters, issue is addressed combing gradient descent auxiliary objective refine efficiency process proved relationship quadratic functions followed justification. In addition, effectiveness proposed validated statistical F measure application. validation results show that significantly trade-off accuracy comparison algorithms under mean squared error (MSE) criterion.