A semi-supervised auto-encoder using label and sparse regularizations for classification

Semi-Supervised Autoencoder for Chemical Gas Classification with FTIR Spectrum

Chemical warfare agents pose a serious threat due to their extreme toxicity, necessitating swift the identification of chemical gases and individual responses to the identified threats. Fourier transform infrared (FTIR) spectroscopy offers a method for remote material analysis, particularly in detecting colorless and odorless chemical agents. In this paper, we propose a deep neural network utilizing a semi-supervised autoencoder (SSAE) for the classification of chemical gases based on FTIR spectra. In contrast to traditional methods, the SSAE concurrently trains an autoencoder and a classifier attached to a latent vector of the autoencoder, enhancing feature extraction for classification. The SSAE was evaluated on laboratory-collected FTIR spectra, demonstrating a superior classification performance compared to existing methods. The efficacy of the SSAE lies in its ability to generate denser cluster distributions in latent vectors, thereby enhancing gas classification. This study established a consistent experimental environment for hyperparameter optimization, offering valuable insights into the influence of latent vectors on classification performance.

Deep Learning for Operating Performance Assessment of Industrial Processes with Layer Attention-Based Stacked Performance-Relevant Denoising Auto-Encoders

The operating performance of a process may degenerate due to process interferences and operation errors, which cancel the benefits of technology design and economic production. Traditional operating performance assessment methods are either lack of real-time due to the post-analysis or difficult to distinguish performance grades for process data with weak differences and strong noise interferences based on shallow learning structures. In this paper, a new layer attention-based stacked performance-relevant denoising auto-encoder (LA-SPDAE) is proposed for the operating performance assessment of industrial processes. It overcomes the defect that the original SDAE ignores task-relevant information in training and only uses the feature of the last hidden layer to complete special tasks. In this study, the original SDAE is improved by optimizing the cross-entropy loss of the performance grade labels in the layer-wise pretraining, which is named stacked performance-relevant denoising auto-encoder (SPDAE), and the performance-relevant features can be extracted under supervision. Moreover, for making good use of performance-relevant features of each layer, they are fused by adaptive weights based on the layer attention mechanism. In the case study of cyanide leaching, the assessment accuracy of the proposed LA-SPDAE model is up to 99.85% under the corrupted proportion of 20%, and the advantage is still maintained as the proportion increases to 80%, which demonstrates the superiority of LA-SPDAE compared with conventional deep neural networks and shallow learning structures.

Automatic classification of nerve discharge rhythms based on sparse auto-encoder and time series feature

Background

Nerve discharge is the carrier of information transmission, which can reveal the basic rules of various nerve activities. Recognition of the nerve discharge rhythm is the key to correctly understand the dynamic behavior of the nervous system. The previous methods for the nerve discharge recognition almost depended on the traditional statistical features, and the nonlinear dynamical features of the discharge activity. The artificial extraction and the empirical judgment of the features were required for the recognition. Thus, these methods suffered from subjective factors and were not conducive to the identification of a large number of discharge rhythms.

Results

The ability of automatic feature extraction along with the development of the neural network has been greatly improved. In this paper, an effective discharge rhythm classification model based on sparse auto-encoder was proposed. The sparse auto-encoder was used to construct the feature learning network. The simulated discharge data from the Chay model and its variants were taken as the input of the network, and the fused features, including the network learning features, covariance and approximate entropy of nerve discharge, were classified by Softmax. The results showed that the accuracy of the classification on the testing data was 87.5%, which could provide more accurate classification results. Compared with other methods for the identification of nerve discharge types, this method could extract the characteristics of nerve discharge rhythm automatically without artificial design, and show a higher accuracy.

Conclusions

The sparse auto-encoder, even neural network has not been used to classify the basic nerve discharge from neither biological experiment data nor model simulation data. The automatic classification method of nerve discharge rhythm based on the sparse auto-encoder in this paper reduced the subjectivity and misjudgment of the artificial feature extraction, saved the time for the comparison with the traditional method, and improved the intelligence of the classification of discharge types. It could further help us to recognize and identify the nerve discharge activities in a new way.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04592-3.

Skeleton-based STIP feature and discriminant sparse coding for human action recognition

Purpose

To find a successful human action recognition system (HAR) for the unmanned environments.

Design/methodology/approach

This paper describes the key technology of an efficient HAR system. In this paper, the advancements for three key steps of the HAR system are presented to improve the accuracy of the existing HAR systems. The key steps are feature extraction, feature descriptor and action classification, which are implemented and analyzed. The usage of the implemented HAR system in the self-driving car is summarized. Finally, the results of the HAR system and other existing action recognition systems are compared.

Findings

This paper exhibits the proposed modification and improvements in the HAR system, namely the skeleton-based spatiotemporal interest points (STIP) feature and the improved discriminative sparse descriptor for the identified feature and the linear action classification.

Research limitations/implications

The experiments are carried out on captured benchmark data sets and need to be analyzed in a real-time environment.

Practical implications

The middleware support between the proposed HAR system and the self-driven car system provides several other challenging opportunities in research.

Social implications

The authors’ work provides the way to go a step ahead in machine vision especially in self-driving cars.

Originality/value

The method for extracting the new feature and constructing an improved discriminative sparse feature descriptor has been introduced.

Multi-focus image fusion method using energy of Laplacian and a deep neural network

Multi-focus image fusion consists in the integration of the focus regions of multiple source images into a single image. At present, there are still some common problems in image fusion methods, such as block artifacts, artificial edges, halo effects, and contrast reduction. To address these problems, a novel, to the best of our knowledge, multi-focus image fusion method using energy of Laplacian and a deep neural network (DNN) is proposed in this paper. The DNN is composed of multiple denoising autoencoders and a classifier. The Laplacian energy operator can effectively extract the focus information of source images, and the trained DNN model can establish a valid mapping relationship between source images and a focus map according to the extracted focus information. First, the Laplacian energy operator is used to perform focus measurement for two source images to obtain the corresponding focus information maps. Then, the sliding window technology is used to sequentially obtain the windows from the corresponding focus information map, and all of the windows are fed back to the trained DNN model to obtain a focus map. After binary segmentation and small region filtering, a final decision map with good consistency is obtained. Finally, according to the weights provided by the final decision map, multiple source images are fused to obtain a final fusion image. Experimental results demonstrate that the proposed fusion method is superior to other existing ones in terms of subjective visual effects and objective quantitative evaluation.