The basic Generative Adversarial Networks (GAN) model is composed of the input vector, generator, and discriminator. Among them, the generator and discriminator are implicit function expressions, usually implemented by deep neural networks. GAN can learn the generative model of any data distribution through adversarial methods with excellent performance. It has been widely applied to different areas since it was proposed in 2014. In this review, we introduced the origin, specific working principle, and development history of GAN, various applications of GAN in digital image processing, Cycle-GAN, and its application in medical imaging analysis, as well as the latest applications of GAN in medical informatics and bioinformatics.

Single-cell RNA-sequencing (scRNA-seq) enables the characterization of transcriptomic profiles at the single-cell resolution with increasingly high throughput. However, it suffers from many sources of technical noises, including insufficient mRNA molecules that lead to excess false zero values, termed dropouts. Computational approaches have been proposed to recover the biologically meaningful expression by borrowing information from similar cells in the observed dataset. However, these methods suffer from oversmoothing and removal of natural cell-to-cell stochasticity in gene expression. Here, we propose the generative adversarial networks (GANs) for scRNA-seq imputation (scIGANs), which uses generated cells rather than observed cells to avoid these limitations and balances the performance between major and rare cell populations. Evaluations based on a variety of simulated and real scRNA-seq datasets show that scIGANs is effective for dropout imputation and enhances various downstream analysis. ScIGANs is robust to small datasets that have very few genes with low expression and/or cell-to-cell variance. ScIGANs works equally well on datasets from different scRNA-seq protocols and is scalable to datasets with over 100 000 cells. We demonstrated in many ways with compelling evidence that scIGANs is not only an application of GANs in omics data but also represents a competing imputation method for the scRNA-seq data.

Single-cell RNA-sequencing (scRNA-seq) enables the characterization of transcriptomic profiles at the single-cell resolution with increasingly high throughput. However, it suffers from many sources of technical noises, including insufficient mRNA molecules that lead to excess false zero values, termed dropouts. Computational approaches have been proposed to recover the biologically meaningful expression by borrowing information from similar cells in the observed dataset. However, these methods suffer from oversmoothing and removal of natural cell-to-cell stochasticity in gene expression. Here, we propose the generative adversarial networks (GANs) for scRNA-seq imputation (scIGANs), which uses generated cells rather than observed cells to avoid these limitations and balances the performance between major and rare cell populations. Evaluations based on a variety of simulated and real scRNA-seq datasets show that scIGANs is effective for dropout imputation and enhances various downstream analysis. ScIGANs is robust to small datasets that have very few genes with low expression and/or cell-to-cell variance. ScIGANs works equally well on datasets from different scRNA-seq protocols and is scalable to datasets with over 100,000 cells. We demonstrated in many ways with compelling evidence that scIGANs is not only an application of GANs in omics data but also represents a competing imputation method for the scRNA-seq data.

Deep convolutional neural networks (CNNs) with strong learning abilities have been used in the field of image super-resolution. However, some CNNs depends on a single deep network to training an image super-resolution model, which will have poor performance in complex screens. To address this problem, we propose a hybrid denoising CNN (HDCNN). HDCNN is composed of a dilated block (DB), RepVGG block (RVB) and feature refinement block (FB), a single convolution. DB combines a dilated convolution, batch normalization (BN), common convolutions, activation function of ReLU to obtain more context information. RVB uses parallel combination of convolution and BN, ReLU to extract complementary width features. FB is used to obtain more accurate information via refining obtained feature from the RVB. A single convolution collaborates a residual learning operation to construct a clean image. These key components make the HDCNN have good performance in image denoising. Experiment shows that the proposed HDCNN enjoys good denoising effect in public datasets.

Fault diagnosis of rotating machinery mainly includes fault feature extraction and fault classification. Vibration signal from the operation of machinery usually could help diagnosing the operational state of equipment. Different types of fault usually have different vibrational features, which are actually the basis of fault diagnosis. This paper proposes a novel fault diagnosis model, which extracts features by combining vibration severity, dyadic wavelet energy time-spectrum, and coefficient power spectrum of the maximum wavelet energy level (VWC) at the feature extraction stage. At the stage of fault classification, we design a support vector machine (SVM) based on the modified shuffled frog-leaping algorithm (MSFLA) for the accurate classifying machinery fault method. Specifically, we use the MSFLA method to optimize SVM parameters. MSFLA can avoid getting trapped into local optimum, speeding up convergence, and improving classification accuracy. Finally, we evaluate our model on real rotating machinery platform, which has four different states, i.e., normal state, eccentric axle fault (EAF), bearing pedestal fault (BPF), and sealing ring wear fault (SRWF). As demonstrated by the results, the VWC method is efficient in extracting vibration signal features of rotating machinery. Based on the extracted features, we further compare our classification method with other three fault classification methods, i.e., backpropagation neural network (BPNN), artificial chemical reaction optimization algorithm (ACROA-SVM), and SFLA-SVM. The experiment results show that MSFLA-SVM achieves a much higher fault classification rate than BPNN, ACROA-SVM, and SFLA-SVM.