21 February 2024 in IEEE Access
Currently, with the widespread application of embedded technology and the continuous improvement of computational power in mobile terminals, the efficient deployment of algorithms on embedded devices, while maintaining high accuracy and minimizing model size, has become a research hotspot. This paper addresses the challenges of deploying the YOLOv8 algorithm on embedded devices and proposes a novel lightweight object detection algorithm focusing on small object detection. We optimize the model through two key strategies, aiming to achieve lightweight deployment and improve the accuracy of small object detection. Firstly, GhostNet is introduced as the backbone network for YOLOv8 in order to achieve lightweight deployment. By using some cost effective operations to generate redundant feature maps, we not only reduce the number of model parameters while ensuring better detection results, but also improve the speed of the model. Secondly, a new multi-scale attention module is designed to enhance the network’s acquisition of crucial information for small targets, which includes a multi-scale fusion attention mechanism and the Soft-NMS algorithm. The multi-scale fusion attention mechanism captures key features of discriminative small targets in the feature map tensor from both spatial and channel dimensions, suppressing non-key information, reducing the impact of complex and unimportant information in the image, enhancing the network model’s learning ability for important features of small targets. The Soft-NMS method improves accuracy by significantly reduces false positives in the detection results. To validate the performance of our proposed method, we conducted validation experiments on the PASCAL VOC dataset and evaluated the model’s generalization ability on the MS COCO dataset. The experiments results demonstrate that our model achieves a significant improvement in small object detection, with a 5.41% increase in detection accuracy compared to the existing YOLOv8. Meanwhile, FLOPs are reduced by 49.62%, and the number of model parameters is reduced by 48.66%. These results fully confirm the effectiveness of our innovative method in achieving both lightweight deployment and significant efficacy in small object detection tasks.
4 December 2017 in IEEE Access
Active learning for networked data that focuses on predicting the labels of other nodes accurately by knowing the labels of a small subset of nodes is attracting more and more researchers because it is very useful especially in cases where labeled data are expensive to obtain. However, most existing researches either only apply to networks with assortative community structure or focus on node attribute data with links or are designed for working in single mode that will work at a higher learning and query cost than batch active learning in general. In view of this, in this paper, we propose a batch mode active learning method which uses information-theoretic techniques and random walk to select which nodes to label. The proposed method requires only network topology as its input, does not need to know the number of blocks in advance and makes no initial assumptions about how the blocks connect. We test our method on two different types of networks: assortative structure and diassortative structure, and then compare our method with a single mode active learning method that is similar to our method except for working in single mode and several simple batch mode active learning methods using information-theoretic techniques and simple heuristics such as employing degree or betweenness centrality. The experimental results show that the proposed method in this paper significantly outperforms them.
14 January 2019 in IEEE Access
The classification of networked data is an interesting and challenging problem. Most traditional relational classifiers that are based on the principle of homophily have an unsatisfactory classification performance in networks with heterophily. This is because these methods treat inhomogeneous networks homogeneously. A progression of a network-only Bayes classifier based secondorder Markov assumption is proposed for heterophilous networks in this paper to address this problem. First, we estimate the class distribution of an unlabeled node according to the class distribution of its neighbors’ neighbors. In this process, we perform this computation on the known and unknown neighbors separately. Second, we combine the two parts using multinomial naïve Bayesian classification. Meanwhile, we pair a relaxation labeling collective inference method (which imports simulated annealing) with this new method to update the class distributions at each iteration. Comparisons of experimental results demonstrate that the proposed method performs better when the networks are heterophilous.
27 February 2020 in IEEE Access
Gene regulatory networks (GRNs) are underlying networks identified by interactive relationships between genes. Reconstructing GRNs from massive genetic data is important for understanding gene functions and biological mechanism, and can provide effective service for medical treatment and genetic research. A series of artificial intelligence based methods have been proposed to infer GRNs from both gene expression data and genetic perturbations. The accuracy of such algorithms can be better than those models that just consider gene expression data. A structural equation model (SEM), which provides a systematic framework integrating both types of gene data conveniently, is a commonly used model for GRN inference. Considering the sparsity of GRNs, in this paper, we develop a novel sparse Bayesian inference algorithm based on Normal-Equation-Gamma (NEG) type hierarchical prior (BaNEG) to infer GRNs modeled with SEMs more accurately. First, we reparameterize an SEM as a linear type model by integrating the endogenous and exogenous variables; Then, a Bayesian adaptive lasso with a three-level NEG prior is applied to deduce the corresponding posterior mode and estimate the parameters. Simulations on synthetic data are run to compare the performance of BaNEG to some state-of-the-art algorithms, the results demonstrate that the proposed algorithm visibly outperforms the others. What's more, BaNEG is applied to infer underlying GRNs from a real data set composed of 47 yeast genes from Saccharomyces cerevisiae to discover potential relationships between genes.
14 January 2019 in IEEE Access
There has been an increasing interest in detecting community in signed networks because signed networks contain more information (both positive and negative edges) than unsigned networks (only positive edges). Many methods have been proposed to find communities in signed networks, however, most of them can be regarded as the discriminative methods that do not concern with how the signed networks are generated so that they are usually difficult to characterize accurately the intrinsic community structure. The existing method SSL which is based on a generative model, can achieve high accuracy in community detection in signed networks. However, SSL needs to estimate the models in the possible space one by one, which requires a large amount of calculation. In view of this, we propose a method to find community in signed networks, in which the exact integrated complete data likelihood (ICLex) for the signed stochastic block model proposed in SSL is derived and a greedy search is employed to optimize the value of the derived ICLex for signed networks to find communities. Our method has a rigorous probabilistic interpretation and does not need to estimate the models one by one from the possible space. The knowledge on hyper-parameters of our model is not necessary. In the experiments, the proposed method is tested on the synthetic and real-world networks and compared with several current methods. Experimental results show that our method can find the communities in signed networks more accurately than these current methods and more efficiently than SSL.