Search

Recent advancements in mobile neural networks, such as the squeeze-and-excitation (SE) attention mechanism, have significantly improved model performance. However, they often overlook the crucial interaction between location information and channels. The interaction of multiple dimensions in feature engineering is of paramount importance for achieving high-quality results. The Transformer model and its successors, such as Mamba and Vision Mamba, have effectively combined features and linked location information. This approach has transitioned from NLP (natural language processing) to CV (computer vision). This paper introduces a novel attention mechanism for mobile neural networks inspired by the structure of Vim (Vision Mamba). It adopts a “1 + 3” architecture to embed multi-dimensional information into channel attention, termed ”Multi-Dimensional Vim-like Attention Mechanism”. The proposed method splits the input into two major branches: the left branch retains the original information for subsequent feature screening, while the right branch divides the channel attention into three one-dimensional feature encoding processes. These processes aggregate features along one channel direction and two spatial directions, simultaneously capturing remote dependencies and preserving precise location information. The resulting feature maps are then combined with the left branch to produce direction-aware, location-sensitive, and channel-aware attention maps. The multi-dimensional Vim-like attention module is simple and can be seamlessly integrated into classical mobile neural networks such as MobileNetV2 and ShuffleNetV2 with minimal computational overhead. Experimental results demonstrate that this attention module adapts well to mobile neural networks with a low parameter count, delivering excellent performance on the CIFAR-100 and MS COCO datasets.

Abstract Single-site pincer-ligated iridium complexes exhibit the ability for C-H activation in homogeneous catalysis. However, instability and difficulty in catalyst recycling are inherent disadvantages of the homogeneous catalyst, limiting its development. Here, we report an atomically dispersed Ir catalyst as the bridge between homogeneous and heterogeneous catalysis, which displays an outstanding catalytic performance for n-butane dehydrogenation, with a remarkable n-butane reaction rate (8.8 mol·gIr −1·h−1) and high butene selectivity (95.6%) at low temperature (450 °C). Significantly, we correlate the BDH activity with the Ir species from nanoscale to sub-nanoscale, to reveal the nature of structure-dependence of catalyst. Moreover, we compare Ir single atoms with Pt single atoms and Pd single atoms for in-depth understanding the nature of metal-dependence at the atomic level. From experimental and theoretical calculations results, the isolated Ir site is suitable for both reactant adsorption/activation and product desorption. Its remarkable dehydrogenation capacity and moderate adsorption behavior are the key to the outstanding catalytic activity and selectivity.

Over the past decade many researchers have applied machine learning algorithms with computational chemistry and materials science tools to explore properties of catalysts. There is a rapid increase in publications demonstrating the use of machine learning for rational catalyst design. In our perspective, targeted tools for rational catalyst design will continue to make significant contributions. However, the community should focus on developing high-throughput simulation tools that utilize molecular dynamics capabilities for thorough exploration of the complex potential energy surfaces that exist, particularly in heterogeneous catalysis. Catalyst-specific databases should be developed to contain enough data to represent the complex multi-dimensional space that defines structure-function relationships. Machine learning tools will continue to impact rational catalyst design; however, we believe that more sophisticated pattern recognition algorithms would yield better understanding of structure-function relationships for heterogeneous catalysis.

The Raman spectroscopy analysis technique has found extensive applications across various disciplines due to its exceptional convenience and efficiency, facilitating the analysis and identification of diverse substances. In recent years, owing to the escalating demand for high-efficiency analytical methods, deep learning models have progressively been introduced into the realm of Raman spectroscopy. However, the application of these models to portable Raman spectrometers has posed a series of challenges due to the computational intensity inherent to deep learning approaches. This paper proposes a lightweight classification model, named RepDwNet, for identifying 28 different types of biological blood. The model integrates advanced techniques such as multi-scale convolutional kernels, depth-wise separable convolutions, and residual connections. These innovations enable the model to capture features at different scales while preserving the coherence of feature data to the maximum extent. The experimental results demonstrate that the average recognition accuracy of the model on the reflective Raman blood dataset and the transmissive Raman blood dataset are 97.31% and 97.10%, respectively. Furthermore, by applying structural reparameterization to compress the well-trained model, it maintains high classification accuracy while significantly reducing the parameter size, thereby enhancing the speed of classification inference. This makes the model more suitable for deployment in portable and mobile devices. Additionally, the proposed model can be extended to various Raman spectroscopy classification scenarios.

With the popularity of mobile devices, lightweight deep learning models have important value in various application scenarios. However, how to effectively fuse the feature information from different dimensions while ensuring the model’s lightness and high accuracy is a problem that has not been fully solved. In this paper, we propose a novel feature fusion module, called φunit, which can fuse the features extracted by different dimensional networks according to the order of feature information with a small computational cost, avoiding the problems of information fragmentation caused by simple feature stacking in traditional information fusion. Based on φunit, this paper further builds an extremely lightweight model φNet, which can achieve performance close to the highest accuracy on several public datasets under the condition of very limited parameter scale. The core idea of φunit is to use deconvolution to reduce the discrepancy among the features to be fused, and to lower the possibility of feature information fragmentation after fusion by fusing the features from different dimensions sequentially. φNet is a lightweight network composed of multiple φunits and bottleneck modules, with a parameter scale of only 1.24 M, much smaller than traditional lightweight models. This paper conducts experiments on public datasets, and φNet achieves an accuracy of 71.64% on the food101 dataset, and an accuracy of 75.31% on the random 50-category food101 dataset, both higher than or close to the highest accuracy. This paper provides a new idea and method for feature fusion of lightweight models, and also provides an efficient model selection for deep learning applications on mobile devices.

Achieving CO oxidation at room temperature is significant for gas purification but remains challenging to perform. Here, the authors report design of graphene-isolated Pt from cobalt-nickel nanoparticles for efficiently catalytic CO oxidation in an oxygen-rich atmosphere.

The role of this study was to evaluate the impact of gut microbiota depletion on the progression of osteoarthritis (QA) and osteoporosis (QP). We conducted an experimental mouse model of QA and QP over an 8-week period. The model involved destabilization of the medial meniscus and bilateral ovariectomy (QVX). To deplete the gut microbiota, we administered a course of antibiotics for 8 weeks. The severity of QA was assessed through micro-CT scanning, X-rays, and immunohistochemical staining. Microbiome analysis was performed using PCR of 16S DNA on fecal samples, and the levels of serum lipopolysaccharide, interleukin 6, tumor necrosis factor-α (TNF-α), osteocalcin, and estrogen were measured using enzyme-linked immunosorbent assay. We found that in comparison to the QVX+QA group, the QVX+QA+ABT group exhibited increased bone mineral density (P < 0.0001), bone volume fraction (P = 0.0051), and trabecular number (P = 0.0023) in the metaphyseal bone. Additionally, cartilage injury and levels of matrix metalloproteinase 13 were reduced in the QVX+QA+ABT group compared to the QVX+QA group. Moreover the QVX+QA+ABT group demonstrated decreased relative abundance of Bacteroidetes, serum lipopolysaccharide (P = 0.0005), TNF-α (P < 0.0001), CTX-1 (P = 0.0002), and increased expression of bone formation markers. These findings were further supported by correlation network analyses. Depletion of gut microbiota was shown to protect against bone loss and cartilage degradation by modulating the composition of the gut microbiota in osteoporosis and osteoarthritis. [ABSTRACT FROM AUTHOR]