Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation.
Yoshiki Narimatsu, Cayetano Pleguezuelos-Manzano, Daniel Hornikx
et al.
Degradation of mucins at the host–microbial mucus interphase involves glycosidases that release monosaccharides from O-glycans and mucinases that cleave the mucin protein backbone. Mucinases recognize and cleave peptide bonds at specific sequence motifs with varying O-glycan structures required and/or permissible. Mucinases that digest mucins with intact O-glycans can potentially destroy the protective mucus, while mucinases that only digest mucins with partially degraded O-glycans may serve at a later stage of nutrient sourcing from mucins. Here, we discovered nine CBM-bearing M60-like mucinases across gut commensals and opportunists, including a conserved Bacteroides fragilis mucinase denoted HC11. We also investigated the previously described Bacteroides thetaiotaomicron mucinase BT4244, which together delineates two functional classes with distinct preferences: BT4244 for bis-Tn (GalNAcα1-O-Ser/Thr) and HC11 for bis-T (Galβ1-3GalNAcα1-O-Ser/Thr) O-glycans. Both mucinases harbor carbohydrate-binding modules (CBM32) that bind their cognate O-glycan motifs and are required – together with the catalytic domains – for efficient cleavage of extended mucin domains, which is consistent with cooperative engagement, but are not required for the cleavage of short glycopeptides. We show B. fragilis strains secrete HC11 and degrade mucins only after the removal of sialic acids. Together, these findings expand the mucinase repertoire by nine enzymes spanning commensals and opportunists, demonstrate that CBM32 domains are essential for efficient cleavage of extended mucin substrates likely by promoting multivalent engagement and substrate positioning, and nominateidentify CBM–catalytic cooperation as a mechanism and intervention point for controlling mucus turnover and barrier integrity.
Diseases of the digestive system. Gastroenterology
Hussein A. Kazem, Miqdam T. Chaichan, Ali H.A. Al-Waeli
et al.
Integrating the photovoltaic/thermal (PV/T) system in green hydrogen production is an improvement in sustainable energy technologies. In PV/T systems, solar energy is converted into electricity and thermal energy simultaneously using hot water or air together with electricity. This dual use saves a significant amount of energy and officially fights greenhouse gases. Different cooling techniques have been proposed in the literature for improving the overall performance of the PV/T systems; employing different types of agents including nanofluids and phase change materials. Hydrogen is the lightest and most abundant element in the universe and has later turned into a flexible energy carrier for transportation and other industrial applications. Issues, including the processes of Hydrogen manufacturing, preservation as well as some risks act as barriers. This paper provides an analysis of several recent publications on the efficiency of using PV/T technology in the process of green hydrogen production and indicates the potential for its increased efficiency as compared to conventional systems that rely on fossil fuels. Due to the effective integration of solar energy, the PV/T system can play an important role in the reduction of the levelized cost of hydrogen (LCOH) and hence play an important part in reducing the economic calculations of the decarbonized energy system.
Energy conservation, Energy industries. Energy policy. Fuel trade
Norah Saleh Alghamdi, Mohammed Zakariah, Achyut Shankar
et al.
Heart disease continues to be a prominent cause of death globally, emphasizing the critical requirement for precise prediction techniques and prompt therapies. This research presents a new method that utilizes the collective capabilities of autoencoder and DenseNet architectures to predict heart illness. Our study is based on the Heart Disease UCI Cleveland dataset, which includes 13 variables that cover clinical and demographic parameters such as age, sex, cholesterol levels, and exercise-induced angina. The dataset presents issues due to its varied attribute types, including category and numerical variables. Furthermore, our approach tackles these difficulties by utilizing a dense autoencoder model, which produced exceptional outcomes. The Model attained a mean accuracy of 99.67% on the Heart Disease UCI Cleveland dataset. Further testing showed it was resilient, with a test accuracy of 99.99%. In addition, the Model demonstrated outstanding macro precision, macro recall, and macro F1 score, with percentages of 99.98%, 99.97%, and 99.96%, respectively. In addition, our results indicate that combining autoencoder and DenseNet designs shows potential for predicting cardiac disease, with substantial enhancements in accuracy and performance metrics compared to current approaches. This methodology can improve clinical decision-making and patient outcomes in cardiovascular care by accurately finding and defining complex patterns within the data. Notwithstanding these encouraging outcomes, our investigation has constraints. The specific attributes of the dataset utilized may limit the applicability of our findings. Subsequent studies could examine the suitability of our method for various datasets and analyze supplementary variables that may improve forecast precision. Furthermore, it is necessary to conduct prospective validation studies to evaluate our strategy’s practical effectiveness in clinical environments.
In light of the low dissolved oxygen concentration in the deep sea, the corrosion mechanisms of the high entropy alloy (HEA) AlCoCrFeNi in artificial seawater with varying oxygen concentrations (2.0, 4.0, 7.0 mg/L) were studied. As the oxygen concentration decreases, the alloy's free corrosion potential decreases, and at 2.0 mg/L, the corrosion rate is 421 times higher than that at 7.0 mg/L. The corrosion form transforms from pitting to uniform corrosion. The primary reasons for this are the passivation film is thin under low oxygen concentration conditions, as well as the preferential dissolution of the alloy elements Al and Ni due to their high activity and “local acidizing” properties, respectively. In designing a super corrosion-resistant high entropy alloy for use in the deep sea, it is advisable to avoid the use of element Al and to add Ni with caution.
Mei-Yu Lv,1,2 Ling-Ling Jin,2,3 Xi-Qiao Sang,2 Wen-Chao Shi,2 Li-Xia Qiang,2 Qing-Yan Lin,4 Shou-De Jin2 1Department of Respiratory Medicine, Harbin Medical University Cancer Hospital, Harbin, 150001, People’s Republic of China; 2Department of Respiratory Medicine, the Fourth Affiliated Hospital of Harbin Medical University, Harbin, 150001, People’s Republic of China; 3Department of Critical Care medicine, the Second Affiliated Hospital of Xi ‘an Jiaotong University, Xi’an, Shaanxi, China; 4Department of Respiratory Medicine, Heilongjiang Provincial Hospital, Harbin, 150001, People’s Republic of ChinaCorrespondence: Shou-De Jin, Department of Respiratory Medicine, The Fourth Affiliated Hospital of Harbin Medical University, No. 37 Yiyuan Street, Nangang District, Harbin, 150001, People’s Republic of China, Tel/Fax +86 0451-85939123, Email jinshoude@163.comPurpose: The typical characteristic of COPD is airway remodeling, affected by environmental and genetic factors. However, genetic studies on COPD have been limited. Currently, the Abhd2 gene is found to play a critical role in maintaining alveolar architecture and stability. The research aims to investigate the predictive value of Abhd2 for airway remodeling in COPD and its effect on TGF-β regulation.Methods: In humans, Abhd2 protein was obtained from peripheral blood monocytes. Peripheral blood TGF-β, pulmonary surfactant proteins (SPs), metalloproteinases, inflammatory indicators (WBC, NEU, NLR, EOS, CRP, PCT, D-Dimer), chest CT (airway diameter and airway wall thickness), pulmonary function, and blood gas analysis were used to assess airway remodeling. In animals, Abhd2 deficient mice (Abhd2Gt/Gt) using gene trapping and C57BL6 mice were injected intraperitoneally with CSE to construct COPD models. HE staining, Masson staining and immunohistochemistry were used to observe the pathological changes of airway in mice, and RT-PCR, WB, ELISA and immunofluorescence were used to detect the expression of secreted proteins and EMT markers.Results: COPD patients with worse pulmonary function and higher airway remodeling-related inflammatory factors had lower Abhd2 protein expression. Moreover, indicators followed the same trend for COPD patients grouped by prognosis (Group A vs Group B). Serum TGF-β was negatively correlated with Abhd2 protein expression, FEV1/FVC, FEV1, and FEV1% PRED. In mice, Abhd2 depletion promoted deposition of TGF-β, leading to more pronounced emphysema, airway thickening, increased alveolar macrophage infiltration, decreased AECII number and SPs, and EMT phenomenon.Conclusion: Downregulation of Abhd2 can promote airway remodeling in COPD by modulating repair after injury and EMT via TGF-β. This study suggests that Abhd2 may serve as a biomarker for assessing airway remodeling and guiding prognosis in COPD.Keywords: alpha/beta hydrolase 2, COPD, TGF-β, airway remodeling, EMT
This paper introduces a router delay model that accurately models key aspects of modern routers. The model accounts for the pipelined nature of contemporary routers, the specific flow control method employed the delay of the flow control credit path, and the sharing of crossbar ports across virtual channels. Motivated by this model, we introduce a microarchitecture for a speculative virtual-channel router that significantly reduces its router latency to that of a brown hole router. Simulations using our pipelined model give results that differ considerably from the commonly assumed 'unit-latency' model which is unreasonably optimistic. Using realistic pipeline models, we compare wormhole and virtual-channel flow control. Our results show that a speculative virtual-channel router has the same per-hop router latency as a wormhole router while improving throughput by up to 40%.
Fatima Rauf, Muhammad Attique Khan, Ali Kashif Bashir
et al.
Despite a worldwide decline in maternal mortality over the past two decades, a significant gap persists between low- and high-income countries, with 94% of maternal mortality concentrated in low and middle-income nations. Ultrasound serves as a prevalent diagnostic tool in prenatal care for monitoring fetal growth and development. Nevertheless, acquiring standard fetal ultrasound planes with accurate anatomical structures proves challenging and time-intensive, even for skilled sonographers. Therefore, for determining common maternal fetuses from ultrasound images, an automated computer-aided diagnostic (CAD) system is required. A new residual bottleneck mechanism-based deep learning architecture has been proposed that includes 82 layers deep. The proposed architecture has added three residual blocks, each including two highway paths and one skip connection. In addition, a convolutional layer has been added of size 3 × 3 before each residual block. In the training process, several hyper parameters have been initialized using Bayesian optimization (BO) rather than manual initialization. Deep features are extracted from the average pooling layer and performed the classification. In the classification process, an increase occurred in the computational time; therefore, we proposed an improved search-based moth flame optimization algorithm for optimal feature selection. The data is then classified using neural network classifiers based on the selected features. The experimental phase involved the analysis of ultrasound images, specifically focusing on fetal brain and common maternal fetal images. The proposed method achieved 78.5% and 79.4% accuracy for brain fetal planes and common maternal fetal planes. Comparison with several pre-trained neural nets and state-of-the-art (SOTA) optimization algorithms shows improved accuracy.