In deep reinforcement learning (RL), learning stochastic and multimodal policies are crucial for various tasks, but most continuous control algorithms model the policy using a deterministic or unimodal Gaussian distribution. Despite being designed to inherently learn a stochastic policy under the maximum entropy RL framework, soft actor-critic (SAC) also uses a factorized Gaussian policy for tractable optimization, which not only restricts the expressiveness of the policy but also ignores correlations among the components of the action vector. In this article, we revisit the approach of employing normalizing flow for SAC policy, justified by the change of variable theorem, and then propose a state-dependent nonvolume preserving (SD-NVP) architecture suitable for SAC learning. In addition, we introduce a generative adversarial SAC (GASAC) that implicitly defines and optimizes various divergences without calculating the normalization constant through a generative adversarial loss. Experimental results on multigoal environment and MuJoCo continuous control tasks suite demonstrate that GASAC model multimodal policy and learn policy more stably in terms of cumulative return.

The advancement of multimedia and artificial intelligence (AI) technologies has dismantled the barriers of information sharing, yet it has also ushered in a double-edged sword: a surge in the spread of fake information. In this context, there is a growing need for research on the detection of 'cheapfakes,' which are low-cost fake media, known for their ease of creation. This paper proposes a multi-stage deep learning process designed to effectively detect the diverse and rapidly evolving nature of cheapfakes. A single-step deep learning model faces limitations in distinguishing various types of cheapfakes, necessitating the application of a complex deep learning model approach to detect subtle Out-of-Context (OOC) phenomena. This study employs models based on Bidirectional Encoder Representations from Transformers (BERT) and stable diffusion technologies to approach cheapfake detection. Through the ACM ICMR 2024 challenge, the performance of this model was evaluated on a real dataset, achieving an accuracy of 71.9% in Task 1, an improvement of 7% over previous methods, and an accuracy of 55.7% in Task 2. These results are expected to make a significant contribution to the development of strategies for creating and countering cheapfakes. Additionally, this research aims to contribute to the detection of OOC media misuse through this challenge.

Recent studies have extensively used deep learning algorithms to analyze gene expression to predict disease diagnosis, treatment effectiveness, and survival outcomes. Survival analysis studies on diseases with high mortality rates, such as cancer, are indispensable. However, deep learning models are plagued by overfitting owing to the limited sample size relative to the large number of genes. Consequently, the latest style-transfer deep generative models have been implemented to generate gene expression data. However, these models are limited in their applicability for clinical purposes because they generate only transcriptomic data. Therefore, this study proposes ctGAN, which enables the combined transformation of gene expression and survival data using a generative adversarial network (GAN). ctGAN improves survival analysis by augmenting data through style transformations between breast cancer and 11 other cancer types. We evaluated the concordance index (C-index) enhancements compared with previous models to demonstrate its superiority. Performance improvements were observed in nine of the 11 cancer types. Moreover, ctGAN outperformed previous models in seven out of the 11 cancer types, with colon adenocarcinoma (COAD) exhibiting the most significant improvement (median C-index increase of ~15.70%). Furthermore, integrating the generated COAD enhanced the log-rank p-value (0.041) compared with using only the real COAD (p-value = 0.797). Based on the data distribution, we demonstrated that the model generated highly plausible data. In clustering evaluation, ctGAN exhibited the highest performance in most cases (89.62%). These findings suggest that ctGAN can be meaningfully utilized to predict disease progression and select personalized treatments in the medical field.

MOTIVATION: Network-based analysis of biomedical data has been extensively studied over the last decades. As a successful application, gene networks have been used to illustrate interactions among genes and explain the associated phenotypes. However, the gene network approaches have not been actively applied for survival analysis, which is one of the main interests of biomedical research. In addition, a few previous studies using gene networks for survival analysis construct networks mainly from prior knowledge, such as pathways, regulations and gene sets, while the performance considerably depends on the selection of prior knowledge. RESULTS: In this paper, we propose a data-driven construction method for survival risk-gene networks as well as a survival risk prediction method using the network structure. The proposed method constructs risk-gene networks with survival-associated genes using penalized regression. Then, gene expression indices are hierarchically adjusted through the networks to reduce the variance intrinsic in datasets. By illustrating risk-gene structure, the proposed method is expected to provide an intuition for the relationship between genes and survival risks. The risk-gene network is applied to a low grade glioma dataset, and produces a hypothesis of the relationship between genetic biomarkers of low and high grade glioma. Moreover, with multiple datasets, we demonstrate that the proposed method shows superior prediction performance compared to other conventional methods. AVAILABILITY AND IMPLEMENTATION: The R package of risk-gene networks is freely available in the web at http://cdal.korea.ac.kr/NetDA/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Motivation: Despite the potential usefulness, the association analysis of gene expression with interval times of two events has been hampered because the occurrence of events can be censored and the conventional survival analysis is not suitable to handle two censored events. However, the recent advances of multivariate survival analysis considering multiple censored events together provide an unprecedented chance for this problem. Based on such advances, we have developed a software tool, GAIT, for the association analysis of gene expression with interval time of two events. Results: The performance of GAIT was demonstrated by simulation studies and the real data analysis. The result indicates the usefulness of GAIT in a wide range of biomedical applications. Availability and implementation: http://cdal.korea.ac.kr/GAIT/index.html. Supplementary information: Supplementary data are available at Bioinformatics online.

Summary: In this paper, we introduce multiple-matching Evidence-based Translator (mEBT) to discover genomic responses from murine expression data for human immune studies, which are significant in the given condition of mice and likely have similar responses in the corresponding condition of human. mEBT is evaluated over multiple datasets and shows improved inter-species agreement. mEBT is expected to be useful for research groups who use murine models to study human immunity. Availability and implementation: http://cdal.korea.ac.kr/mebt/. Supplementary information: Supplementary data are available at Bioinformatics online.

SUMMARY: High-throughput genome-wide studies of alternatively spliced mRNA transcripts have become increasingly important in clinical research. Consequently, easy-to-use software tools are required to process data from these studies, for example, using exon and junction arrays. Here, we introduce JETTA, an integrated software package for the calculation of gene expression indices as well as the identification and visualization of alternative splicing events. We demonstrate the software using data of human liver and muscle samples hybridized on an exon-junction array. AVAILABILITY: JETTA and its demonstrations are freely available at http://igenomed.stanford.edu/~junhee/JETTA/index.html

BACKGROUND: The large amount of high-throughput genomic data has facilitated the discovery of the regulatory relationships between transcription factors and their target genes. While early methods for discovery of transcriptional regulation relationships from microarray data often focused on the high-throughput experimental data alone, more recent approaches have explored the integration of external knowledge bases of gene interactions. RESULTS: In this work, we develop an algorithm that provides improved performance in the prediction of transcriptional regulatory relationships by supplementing the analysis of microarray data with a new method of integrating information from an existing knowledge base. Using a well-known dataset of yeast microarrays and the Yeast Proteome Database, a comprehensive collection of known information of yeast genes, we show that knowledge-based predictions demonstrate better sensitivity and specificity in inferring new transcriptional interactions than predictions from microarray data alone. We also show that comprehensive, direct and high-quality knowledge bases provide better prediction performance. Comparison of our results with ChIP-chip data and growth fitness data suggests that our predicted genome-wide regulatory pairs in yeast are reasonable candidates for follow-up biological verification. CONCLUSION: High quality, comprehensive, and direct knowledge bases, when combined with appropriate bioinformatic algorithms, can significantly improve the discovery of gene regulatory relationships from high throughput gene expression data.