Supervised Dimensionality Reduction Methods via Recursive Regression

A Novel Formulation of Trace Ratio Linear Discriminant Analysis

The linear discriminant analysis (LDA) method needs to be transformed into another form to acquire an approximate closed-form solution, which could lead to the error between the approximate solution and the true value. Furthermore, the sensitivity of dimensionality reduction (DR) methods to subspace dimensionality cannot be eliminated. In this article, a new formulation of trace ratio LDA (TRLDA) is proposed, which has an optimal solution of LDA. When solving the projection matrix, the TRLDA method given by us is transformed into a quadratic problem with regard to the Stiefel manifold. In addition, we propose a new trace difference problem named optimal dimensionality linear discriminant analysis (ODLDA) to determine the optimal subspace dimension. The nonmonotonicity of ODLDA guarantees the existence of optimal subspace dimensionality. Both the two approaches have achieved efficient DR on several data sets.

A Smoothed Matrix Multivariate Elliptical Distribution-Based Projection Method for Feature Extraction

Big data has the traits such as “the curse of dimensionality,” high storage cost, and heavy computation burden. Self-representation-based feature extraction methods cannot effectively deal with the image-level structural noise in the data, so how to character a better relationship of reconstruction representation is very important. Recently, sparse representation with smoothed matrix multivariate elliptical distribution (SMED) using structural information to handle low-rank error images caused by illumination or occlusion has been proposed. Based on SMED, we present a new method named SMEDP for feature extraction. SMEDP firstly utilizes SMED to automatically construct an adjacency graph and then obtains an optimal projection matrix by maximizing the ratio of the local scatter matrix and the total scatter matrix in the PCA subspace. Experiments on the COIL-20 object database, ORL face database, and CMU PIE face database prove that SMEDP works well and can achieve considerable visual and recognition performance than the relevant methods.

Fast Unsupervised Projection for Large-Scale Data

Dimensionality reduction (DR) technique has been frequently used to alleviate information redundancy and reduce computational complexity. Traditional DR methods generally are inability to deal with nonlinear data and have high computational complexity. To cope with the problems, we propose a fast unsupervised projection (FUP) method. The simplified graph of FUP is constructed by samples and representative points, where the number of the representative points selected through iterative optimization is less than that of samples. By generating the presented graph, it is proved that large-scale data can be projected faster in numerous scenarios. Thereafter, the orthogonality FUP (OFUP) method is proposed to ensure the orthogonality of projection matrix. Specifically, the OFUP method is proved to be equivalent to PCA upon certain parameter setting. Experimental results on benchmark data sets show the effectiveness in retaining the essential information.

Labeled-Robust Regression: Simultaneous Data Recovery and Classification

Rank minimization is widely used to extract low-dimensional subspaces. As a convex relaxation of the rank minimization, the problem of nuclear norm minimization has been attracting widespread attention. However, the standard nuclear norm minimization usually results in overcompression of data in all subspaces and eliminates the discrimination information between different categories of data. To overcome these drawbacks, in this article, we introduce the label information into the nuclear norm minimization problem and propose a labeled-robust principal component analysis (L-RPCA) to realize nuclear norm minimization on multisubspace data. Compared with the standard nuclear norm minimization, our method can effectively utilize the discriminant information in multisubspace rank minimization and avoid excessive elimination of local information and multisubspace characteristics of the data. Then, an effective labeled-robust regression (L-RR) method is proposed to simultaneously recover the data and labels of the observed data. Experiments on real datasets show that our proposed methods are superior to other state-of-the-art methods.

Global Trends and Hotspots in Research on Rehabilitation Robots: A Bibliometric Analysis From 2010 to 2020

Background: In recent years, with the development of medical science and artificial intelligence, research on rehabilitation robots has gained more and more attention, for nearly 10 years in the Web of Science database by journal of rehabilitation robot-related research literature analysis, to parse and track rehabilitation robot research hotspot and front, and provide some guidance for future research.

Methods: This study employed computer retrieval of rehabilitation robot-related research published in the core data collection of the Web of Science database from 2010 to 2020, using CiteSpace 5.7 visualization software. The hotspots and frontiers of rehabilitation robot research are analyzed from the aspects of high-influence countries or regions, institutions, authors, high-frequency keywords, and emergent words.

Results: A total of 3,194 articles were included. In recent years, the research on rehabilitation robots has been continuously hot, and the annual publication of relevant literature has shown a trend of steady growth. The United States ranked first with 819 papers, and China ranked second with 603 papers. Northwestern University ranked first with 161 publications. R. Riener, a professor at the University of Zurich, Switzerland, ranked as the first author with 48 articles. The Journal of Neural Engineering and Rehabilitation has the most published research, with 211 publications. In the past 10 years, research has focused on intelligent control, task analysis, and the learning, performance, and reliability of rehabilitation robots to realize the natural and precise interaction between humans and machines. Research on neural rehabilitation robots, brain–computer interface, virtual reality, flexible wearables, task analysis, and exoskeletons has attracted more and more attention.

Conclusions: At present, the brain–computer interface, virtual reality, flexible wearables, task analysis, and exoskeleton rehabilitation robots are the research trends and hotspots. Future research should focus on the application of machine learning (ML), dimensionality reduction, and feature engineering technologies in the research and development of rehabilitation robots to improve the speed and accuracy of algorithms. To achieve wide application and commercialization, future rehabilitation robots should also develop toward mass production and low cost. We should pay attention to the functional needs of patients, strengthen multidisciplinary communication and cooperation, and promote rehabilitation robots to better serve the rehabilitation medical field.

Fast Multiscale Neighbor Embedding

Dimension reduction (DR) computes faithful low-dimensional (LD) representations of high-dimensional (HD) data. Outstanding performances are achieved by recent neighbor embedding (NE) algorithms such as t -SNE, which mitigate the curse of dimensionality. The single-scale or multiscale nature of NE schemes drives the HD neighborhood preservation in the LD space (LDS). While single-scale methods focus on single-sized neighborhoods through the concept of perplexity, multiscale ones preserve neighborhoods in a broader range of sizes and account for the global HD organization to define the LDS. For both single-scale and multiscale methods, however, their time complexity in the number of samples is unaffordable for big data sets. Single-scale methods can be accelerated by relying on the inherent sparsity of the HD similarities they involve. On the other hand, the dense structure of the multiscale HD similarities prevents developing fast multiscale schemes in a similar way. This article addresses this difficulty by designing randomized accelerations of the multiscale methods. To account for all levels of interactions, the HD data are first subsampled at different scales, enabling to identify small and relevant neighbor sets for each data point thanks to vantage-point trees. Afterward, these sets are employed with a Barnes-Hut algorithm to cheaply evaluate the considered cost function and its gradient, enabling large-scale use of multiscale NE schemes. Extensive experiments demonstrate that the proposed accelerations are, statistically significantly, both faster than the original multiscale methods by orders of magnitude, and better preserving the HD neighborhoods than state-of-the-art single-scale schemes, leading to high-quality LD embeddings. Public codes are freely available at https://github.com/cdebodt.

Simpler and Faster Development of Tumor Phylogeny Pipelines

In the recent years, there has been an increasing amount of single-cell sequencing studies, producing a considerable number of new data sets. This has particularly affected the field of cancer analysis, where more and more articles are published using this sequencing technique that allows for capturing more detailed information regarding the specific genetic mutations on each individually sampled cell. As the amount of information increases, it is necessary to have more sophisticated and rapid tools for analyzing the samples. To this goal, we developed plastic (PipeLine Amalgamating Single-cell Tree Inference Components), an easy-to-use and quick to adapt pipeline that integrates three different steps: (1) to simplify the input data, (2) to infer tumor phylogenies, and (3) to compare the phylogenies. We have created a pipeline submodule for each of those steps and developed new in-memory data structures that allow for easy and transparent sharing of the information across the tools implementing the above steps. While we use existing open source tools for those steps, we have extended the tool used for simplifying the input data, incorporating two machine learning procedures-which greatly reduce the running time without affecting the quality of the downstream analysis. Moreover, we have introduced the capability of producing some plots to quickly visualize results.