No abstract

Since tumor is seriously harmful to human health, effective diagnosis measures are in urgent need for tumor therapy. Early detection of tumor is particularly important for better treatment of patients. A notable issue is how to effectively discriminate tumor samples from normal ones. Many classification methods, such as Support Vector Machines (SVMs), have been proposed for tumor classification. Recently, deep learning has achieved satisfactory performance in the classification task of many areas. However, the application of deep learning is rare in tumor classification due to insufficient training samples of gene expression data. In this paper, a Sample Expansion method is proposed to address the problem. Inspired by the idea of Denoising Autoencoder (DAE), a large number of samples are obtained by randomly cleaning partially corrupted input many times. The expanded samples can not only maintain the merits of corrupted data in DAE but also deal with the problem of insufficient training samples of gene expression data to a certain extent. Since Stacked Autoencoder (SAE) and Convolutional Neural Network (CNN) models show excellent performance in classification task, the applicability of SAE and 1-dimensional CNN (1DCNN) on gene expression data is analyzed. Finally, two deep learning models, Sample Expansion-Based SAE (SESAE) and Sample Expansion-Based 1DCNN (SE1DCNN), are designed to carry out tumor gene expression data classification by using the expanded samples. Experimental studies indicate that SESAE and SE1DCNN are very effective in tumor classification.

Recently, deep learning-based methods have drawn increasing attention in hyperspectral imagery (HSI) classification, due to their strong nonlinear mapping capability. However, these methods suffer from a time-consuming training process because of many network parameters. In this paper, the concept of broad learning is introduced into HSI classification. Firstly, to make full use of abundant spectral and spatial information of hyperspectral imagery, hierarchical guidance filtering is performing on the original HSI to get its spectral-spatial representation. Then, the class-probability structure is incorporated into the broad learning model to obtain a semi-supervised broad learning version, so that limited labeled samples and many unlabeled samples can be utilized simultaneously. Finally, the connecting weights of broad structure can be easily computed through the ridge regression approximation. Experimental results on three popular hyperspectral imagery datasets demonstrate that the proposed method can achieve better performance than deep learning-based methods and conventional classifiers.

Extreme Learning Machine (ELM) as a fast and efficient neural network model in pattern recognition and machine learning will decline when the labeled training sample is insufficient. Transfer learning helps the target task to learn a reliable model by using plentiful labeled samples from the different but relevant domain. In this paper, we propose a supervised Extreme Learning Machine with knowledge transferability, called Transfer Extreme Learning Machine with Output Weight Alignment (TELM-OWA). Firstly, it reduces the distribution difference between domains by aligning the output weight matrix of the ELM trained by the labeled samples from the source and target domains. Secondly, the approximation between the interdomain ELM output weight matrix is added to the objective function to further realize the cross-domain transfer of knowledge. Thirdly, we consider the objective function as the least square problem and transform it into a standard ELM model to be efficiently solved. Finally, the effectiveness of the proposed algorithm is verified by classification experiments on 16 sets of image datasets and 6 sets of text datasets, and the result demonstrates the competitive performance of our method with respect to other ELM models and transfer learning approach.

Unsupervised Domain Adaptation (UDA) aims to leverage the knowledge from the labeled source domain to help the task of target domain with the unlabeled data. It is a key step for UDA to minimize the cross-domain distribution divergence. In this paper, we firstly propose a novel discrepancy metric, referred to as Cross Domain Mean Approximation (CDMA) discrepancy, to evaluate the distribution differences between source and target domains, which calculate the sum of the squares of the distances from the source and target domains to the mean of the other domain. Secondly, Joint Distribution Adaptation based on Cross Domain Mean Approximation (JDA-CDMA) is developed on the basis of CDMA to extract shared feature and simultaneously reduce the marginal and conditional distribution discrepancy between domains during the label refinement process. Thirdly, we construct a classifier utilizing CDMA metric and neighbor information. Finally, the proposed feature extraction approach and classifier are combined to realize transfer learning. Results from extensive experiments on five visual benchmarks including object, face, and digit images, show the proposed methods outperform the state-of-the-art unsupervised domain adaptation. I. INTRODUCTION I N machine vision, many machine learning methods, such as Linear Regression [1], Logistic Regression (LR) [2], k-Nearest Neighbor (k-NN) [3], Bayesian [4], Decision Tree [5], and Support Vector Machine (SVM)[6], are applied to image classification tasks. However, when the image feature representation is too redundant or poor in quality, their accuracy will be lowered. Therefore, it is of great importance to extract high-quality image feature. Feature extraction, as an important manner of mining image latent knowledge, is not only conducive to the in-depth understanding of image content, but also crucial to improve the accuracy of image classification and recognition [7]. Consequently it has attracted much attention from researchers. Principal Component Analysis (PCA) [8], Independent Component Analysis (ICA) [9], Linear Discriminant Analysis (LDA) [10], Maximum Margin Criterion (MMC) [11] and other algorithms are often used for feature extraction. In order to discover the nonlinear structure hidden in the high dimensional data and mine the local geometric structure information of data, Laplacian Eigen-maps (LE) [12], Locality Linear Embedding

Deep reinforcement learning (RL) comprehensively uses the psychological mechanisms of "trial and error" and "reward and punishment" in RL as well as powerful feature expression and nonlinear mapping in deep learning. Currently, it plays an essential role in the fields of artificial intelligence and machine learning. Since an RL agent needs to constantly interact with its surroundings, the deep Q network (DQN) is inevitably faced with the need to learn numerous network parameters, which results in low learning efficiency. In this paper, a multisource transfer double DQN (MTDDQN) based on actor learning is proposed. The transfer learning technique is integrated with deep RL to make the RL agent collect, summarize, and transfer action knowledge, including policy mimic and feature regression, to the training of related tasks. There exists action overestimation in DQN, i.e., the lower probability limit of action corresponding to the maximum Q value is nonzero. Therefore, the transfer network is trained by using double DQN to eliminate the error accumulation caused by action overestimation. In addition, to avoid negative transfer, i.e., to ensure strong correlations between source and target tasks, a multisource transfer learning mechanism is applied. The Atari2600 game is tested on the arcade learning environment platform to evaluate the feasibility and performance of MTDDQN by comparing it with some mainstream approaches, such as DQN and double DQN. Experiments prove that MTDDQN achieves not only human-like actor learning transfer capability, but also the desired learning efficiency and testing accuracy on target task.

No abstract

No abstract