Search

With the increase in the performance of deep learning models, the model parameter has increased exponentially. An increase in model parameters leads to an increase in computation and training time, i.e., an increase in training cost. To reduce the training cost, we propose Compositional Intelligence (CI). This is a reuse method that combines pre-trained models for different tasks. Since the CI uses a well-trained model, good performance and small training cost can be expected in the target task. We applied the CI to the Image Captioning task. Compared to using a trained feature extractor, the caption generator is usually trained from scratch. On the other hand, we pre-trained the Transformer model as a caption generator and applied CI, i.e., we used a pre-trained feature extractor and a pre-trained caption generator. To compare the training cost of the From Scratch model and the CI model, early stopping was applied during fine-tuning of the image captioning task. On the MS-COCO dataset, the vanilla image captioning model reduced training cost by 13.8% and improved performance by up to 3.2%, and the Object Relation Transformer model reduced training cost by 21.3%.

Inflammatory cytokines mediate inflammatory bowel diseases (IBDs) and cytokine blocking therapies often ameliorate the disease severity. IL-32 affects inflammation by increasing the production of IL-1, TNFα, and several chemokines. Here, we investigated the role of IL-32 in intestinal inflammation by generating a transgenic (TG) mouse expressing human IL-32γ (IL-32γ TG). Although IL-32γ TG mice are healthy, constitutive serum and colonic tissue levels of TNFα are elevated. Compared with wild-type (WT) mice, IL-32γ TG mice exhibited a modestly exacerbated acute inflammation early following the initiation of dextran sodium sulfate (DSS)-induced colitis. However, after 6 d, there was less colonic inflammation, reduced tissue loss, and improved survival rate compared with WT mice. Associated with attenuated tissue damage, colonic levels of TNFα and IL-6 were significantly reduced in the IL-32γ TG mice whereas IL-10 was elevated. Cultured colon explants from IL-32γ TG mice secreted higher levels of IL-10 compared with WT mice and lower levels of TNFα and IL-6. Constitutive levels of IL-32γ itself in colonic tissues were significantly lower following DSS colitis. Although the highest level of serum IL-32γ occurred on day 3 of colitis, IL-32 was below constitutive levels on day 9. The ability of IL-32γ to increase constitutive IL-10 likely reduces TNFα, IL-6, and IL-32 itself accounting for less inflammation. In humans with ulcerative colitis (UC), serum IL-32 is elevated and colonic biopsies contain IL-32 in inflamed tissues but not in uninvolved tissues. Thus IL-32γ emerges as an example of how innate inflammation worsens as well as protects intestinal integrity.

Proteinase 3 (PR3), a neutrophil granule serine protease, is the major autoantigen for autoantibodies in the systemic vasculitic disease, Wegener's granulomatosis. It is also found to be involved in various inflammatory diseases including Crohn's disease, rheumatoid arthritis, cystic fibrosis, and gingivitis. However, there is no high quality antibody available to detect endogenous PR3 in biological samples such as plasma and tissue. Several commercial anti-PR3 monoclonal antibodies (MAbs) were obtained by using HMC-1/PR3 cell granule extracts as the antigen, but the resulting antibodies could not be applied for immunoblotting or other immunological methods. Therefore, we produced human recombinant PR3 in Escherichia coli and developed several MAbs that are highly sensitive and can be used for immunoblotting, FACS analysis, and immunofluorescent staining. The PR3 MAbs recognized both rhPR3 and human plasma-derived neutrophil PR3 in reducing and non-reducing conditions at low nanogram levels. In addition, new MAbs detect endogenous PR3 from normal human plasma and urine with high specificity. The new anti-PR3 MAbs will be an essential tool for investigating the role of PR3 in inflammatory and autoimmune diseases.

A networked viewpoint controller exploits a spatiotemporal attention module for the smart user interaction of the future 3D TV. In this paper, a new approach to locate a focus of attention for the generation of candidate viewpoints is suggested. The suggested method combines spatial and temporal features given from a series of images to provide viewers with several viewpoints.