Learning optimally separated class-specific subspace representations using convolutional autoencoder
This work addresses the challenge of improving classification accuracy for machine learning tasks where data lies in overlapping subspaces, representing an incremental advancement in subspace-based transformation learning.
The paper tackles the problem of noisy and poorly separated class-specific data in low-dimensional linear subspaces by proposing a convolutional autoencoder with a class-specific self expressiveness layer to generate well-separated subspace representations, resulting in significant improvement in classification performance over existing subspace-based methods.
In this work, we propose a novel convolutional autoencoder based architecture to generate subspace specific feature representations that are best suited for classification task. The class-specific data is assumed to lie in low dimensional linear subspaces, which could be noisy and not well separated, i.e., subspace distance (principal angle) between two classes is very low. The proposed network uses a novel class-specific self expressiveness (CSSE) layer sandwiched between encoder and decoder networks to generate class-wise subspace representations which are well separated. The CSSE layer along with encoder/ decoder are trained in such a way that data still lies in subspaces in the feature space with minimum principal angle much higher than that of the input space. To demonstrate the effectiveness of the proposed approach, several experiments have been carried out on state-of-the-art machine learning datasets and a significant improvement in classification performance is observed over existing subspace based transformation learning methods.