The fusion of multi-band images including far-infrared image (FIRI), near-infrared image (NIRI) and visible image (VISI) primarily suffers from four challenges. One is the problem of simultaneous fusion for multiple images. Most existing methods are oriented towards the fusion of two objects, which is generally achieved with a sequential fusion method. This means that intermediate fusion results are repeatedly integrated with the unprocessed images until all images have been fused. However, this may amplify the blurring effect and even engender artifacts. Second, consistent training labels for image fusion cannot currently be obtained for some types of images (e.g., medical images and multi-band images), which may lead to the failed application of supervised learning methods. Third, the existing methods often do not directly focus on the potential mapping relationship between the original and resulting images, which usually increases the unpredictability of the fusion results. Fourth, redundant features or singularities are often not eliminated in the general fusion process, and both may interfere with or even obscure significant features in the source images. To address the abovementioned problems, this paper proposes a latent representation learning model that can synchronously integrate multi-band images without samples. Specifically, the model can capture the clean and distinctive features of the originals via latent low-rank and sparse embedding. The extracted intrinsic features are projected onto the target fusion space through an assumed mapping relationship. The final results were obtained through the designed optimization algorithm. In addition, numerous experiments were implemented to prove the rationality and feasibility of the proposed fusion model with subjective evaluation, objective indexes and convergence analysis.

Smoking is the major cause of lung cancer and the leading cause of cancer-related death in the world. The most current view about lung cancer is no longer limited to individual genes being mutated by any carcinogenic insults from smoking. Instead, tumorigenesis is a phenotype conferred by many systematic and global alterations, leading to extensive heterogeneity and variation for both the genotypes and phenotypes of individual cancer cells. Thus, strategically it is foremost important to develop a methodology to capture any consistent and global alterations presumably shared by most of the cancerous cells for a given population. This is particularly true that almost all of the data collected from solid cancers (including lung cancers) are usually distant apart over a large span of temporal or even spatial contexts. Here, we report a multiple non-Gaussian graphical model to reconstruct the gene interaction network using two previously published gene expression datasets. Our graphical model aims to selectively detect gross structural changes at the level of gene interaction networks. Our methodology is extensively validated, demonstrating good robustness, as well as the selectivity and specificity expected based on our biological insights. In summary, gene regulatory networks are still relatively stable during presumably the early stage of neoplastic transformation. But drastic structural differences can be found between lung cancer and its normal control, including the gain of functional modules for cellular proliferations such as EGFR and PDGFRA, as well as the lost of the important IL6 module, supporting their roles as potential drug targets. Interestingly, our method can also detect early modular changes, with the ALDH3A1 and its associated interactions being strongly implicated as a potential early marker, whose activations appear to alter LCN2 module as well as its interactions with the important TP53-MDM2 circuitry. Our strategy using the graphical model to reconstruct gene interaction work with biologically-inspired constraints exemplifies the importance and beauty of biology in developing any bio-computational approach.

This study proposes a new deep neural network to solve the multi-band image synchronous fusion problem (MBF-Net). Unlike other deep learning-based methods, our network architecture design combines the ideas of model-driven and data-driven methods, so it is more interpretable. First, a new multi-band image synchronous fusion model is proposed. The source image in the data fidelity terms and the prior regularization are implicitly represented by the deep learning network and jointly learned from the training data. The proposed model is then solved using a half quadratic splitting (HQS) algorithm and unfolded into a deep fusion network. In addition, a new saliency loss function is proposed to retain thermal radiation information to enhance the fusion effect. Finally, the experimental results on the TNO dataset demonstrated the effectiveness of the proposed MBF-Net.

Acoustic signal classification plays a central role in acoustic source identification. In practical applications, however, varieties of training data are typically inadequate, which leads to a low sample complexity. Applying classical deep learning methods to identify acoustic signals involves a large number of parameters in the classification model, which calls for great sample complexity. Therefore, low sample complexity modeling is one of the most important issues related to the performance of the acoustic signal classification. In this study, the authors propose a novel data fusion model named MFF-ResNet, in which manual design features and deep representation of log-Mel spectrogram features are fused with bi-level attention. The proposed approach involves an amount of prior human knowledge as implicit regularization, thus leading to an interpretable and low sample complexity model of the acoustic signal classification. The experimental results suggested that MFF-ResNet is capable of accurate acoustic signal classification with fewer training samples.

Cancer is a heterogeneous disease, thus one of the central problems is how to dissect the resulting complex phenotypes in terms of their biological building blocks. Computationally, this is to represent and interpret high dimensional observations through a structural and conceptual abstraction into the most influential determinants underlying the problem. The working hypothesis of this report is to consider gene interaction to be largely responsible for the manifestation of complex cancer phenotypes, thus where the representation is to be conceptualized. Here we report a representation learning strategy combined with regularizations, in which gene expressions are described in terms of a regularized product of meta-genes and their expression levels. The meta-genes are constrained by gene interactions thus representing their original topological contexts. The expression levels are supervised by their conditional dependencies among the observations thus providing a cluster-specific constraint. We obtain both of these structural constraints using a node-based graphical model. Our representation allows the selection of more influential modules, thus implicating their possible roles in neoplastic transformations. We validate our representation strategy by its robust recognitions of various cancer phenotypes comparing with various classical methods. The modules discovered are either shared or specify for different types or stages of human cancers, all of which are consistent with literature and biology.