Ping Li

Time-series data is an appealing study topic in data mining and has a broad range of applications. Many approaches have been employed to handle time series classification (TSC) challenges with promising results, among which deep neural network methods have become mainstream. Echo State Networks (ESN) and Convolutional Neural Networks (CNN) are commonly utilized as deep neural network methods in TSC research. However, ESN and CNN can only extract local dependencies relations of time series, resulting in long-term temporal data dependence needing to be more challenging to capture. As a result, an encoder and decoder architecture named LA-ESN is proposed for TSC tasks. In LA-ESN, the encoder is composed of ESN, which is utilized to obtain the time series matrix representation. Meanwhile, the decoder consists of a one-dimensional CNN (1D CNN), a Long Short-Term Memory network (LSTM) and an Attention Mechanism (AM), which can extract local information and global dependencies from the representation. Finally, many comparative experimental studies were conducted on 128 univariate datasets from different domains, and three evaluation metrics including classification accuracy, mean error and mean rank were exploited to evaluate the performance. In comparison to other approaches, LA-ESN produced good results.

In this paper, we present a novel unsupervised feature selection method termed robust matrix factorization with robust adaptive structure learning (RMFRASL), which can select discriminative features from a large amount of multimedia data to improve the performance of classification and clustering tasks. RMFRASL integrates three models (robust matrix factorization, adaptive structure learning, and structure regularization) into a unified framework. More specifically, a robust matrix factorization-based feature selection (RMFFS) model is proposed by introducing an indicator matrix to measure the importance of features, and the L₂₁-norm is adopted as a metric to enhance the robustness of feature selection. Furthermore, a robust adaptive structure learning (RASL) model based on the self-representation capability of the samples is designed to discover the geometric structure relationships of original data. Lastly, a structure regularization (SR) term is designed on the learned graph structure, which constrains the selected features to preserve the structure information in the selected feature space. To solve the objective function of our proposed RMFRASL, an iterative optimization algorithm is proposed. By comparing our method with some state-of-the-art unsupervised feature selection approaches on several publicly available databases, the advantage of the proposed RMFRASL is demonstrated.