Jiao Shi’s research while affiliated with Northwestern Polytechnical University and other places

Unsupervised domain adaptation (UDA) is widely utilized to accomplish the visual object recognition task in the field of consumer electronics, aiming at reducing labeling costs in new scenarios. To address the error accumulation problem caused by unreliable target pseudo labels, existing feature-based UDA methods utilize the predicted posterior probability of target sample to model the uncertainty relationship between samples. However, due to differences in data characteristics of each class, unified uncertainty modeling may lead to information loss in some classes due to excessive fuzziness. In view of this, this paper proposes a uncertainty feature learning method with personalized class description for UDA (UFL-PCD) to accomplish the visual object recognition task. Firstly, to learn a more robust discriminative common feature subspace, the proposed method performs more flexible uncertainty modeling of relationships between intra-domain and cross-domain samples based on the personalized uncertainty description of each class. Secondly, to alleviate the error accumulation problem, a hybrid label filtering strategy is proposed to determine the credible target pseudo labels based on fuzzy membership degrees and adaptive rough approximations. Furthermore, a consensus prototype-driven prediction model is designed to predict target data by comprehensively considering the guidance information of the source domain and the intrinsic structural information of the target domain. Experiments conducted on four benchmark datasets have demonstrated that UFL-PCD can construct a common feature subspace with strong stability and adaptability through uncertainty feature learning.

To address the problem of enormous differences in two heterogeneous images, the traditional unsupervised frameworks are most normally realized by converting two images into a common domain with various auxiliary strategies, such as transformation and alignment, which requires extensive calculation and has difficulty balancing the training tasks. To achieve a concise framework, this paper proposes self-guided autoencoders (SGAE) for unsupervised change detection in heterogeneous remote sensing images. Unlike traditional methods that aim to narrow the differences of heterogeneous images to highlight the changed information, SGAE forces the flow of identification in formation generated in unlabeled data through self-guided iterations. First, initial unsupervised networks output an elementary change map that will be screened to obtain reliable pseudolabels. The selected pseudo-labeled samples will be used as the input of a supervised network to obtain another change map. Then, multiple change maps will be fused to refine the confidence of pseudolabels again, obtaining new fused pseudolabeled samples for the self-guided network, which will be trained with pseudo-labeled samples and unlabeled samples. Finally, the above operations will be repeated to continuously optimize the net, which helps itself to extract the discriminative features for classification in self-guided iterations. Experiments compared with several algorithms on four datasets demonstrate the effectiveness and robustness of our method, which can help unsupervised models improve discriminative feature extraction and classification performance with a more flexible learning method.

In recent years, deep learning has achieved remarkable success in classifying hyperspectral images (HSIs), relying heavily on the quantity and quality of labeled samples. However, obtaining sufficient labels for HSIs poses a challenge. HSIs obtained by the same sensor often exhibit similar spectral information due to their shared physical, chemical properties, or reflective attributes. Joint analysis of several HSIs enables the integration of limited labeled samples and extraction of more robust and discriminative features from different HSIs. Therefore, a collaborative network (SSCN) for the joint classification of multiple HSIs acquired by the same sensor in different areas is proposed. In the SSCN, each HSI has its own feature extraction channel, which facilitates the learning of image-specific representations. Additionally, a feature sharing channel is created to extract and transfer multi-hierarchical image-shared representations between multiple HSIs, thereby forming a common knowledge pool to facilitate feature sharing. Furthermore, a cross-channel mutual attention module is designed to collaboratively utilize features from image-specific and image-shared channels, enhancing the efficiency of information communication in HSIs. The experimental results on six HSIs demonstrate that the proposed SSCN can jointly classify multiple HSIs by the same sensor in different areas and achieve good classification performance.

In the absence of sufficient labels, deep neural networks (DNNs) are prone to overfitting, resulting in poor performance and difficulty in training. Thus, many semisupervised methods aim to use unlabeled sample information to compensate for the lack of label quantity. However, as the available pseudolabels increase, the fixed structure of traditional models has difficulty in matching them, limiting their effectiveness. Therefore, a deep-growing neural network with manifold constraints (DGNN-MC) is proposed. It can deepen the corresponding network structure with the expansion of a high-quality pseudolabel pool and preserve the local structure between the original and high-dimensional data in semisupervised learning. First, the framework filters the output of the shallow network to obtain pseudolabeled samples with high confidence and adds them to the original training set to form a new pseudolabeled training set. Second, according to the size of the new training set, it increases the depth of the layers to obtain a deeper network and conducts the training. Finally, it obtains new pseudolabeled samples and deepens the layers again until the network growth is completed. The growing model proposed in this article can be applied to other multilayer networks, as their depth can be transformed. Taking HSI classification as an example, a natural semisupervised problem, the experimental results demonstrate the superiority and effectiveness of our method, which can mine more reliable information for better utilization and fully balance the growing amount of labeled data and network learning ability.

Neural network models for hyperspectral images classification are complex and therefore difficult to deploy directly onto mobile platforms. Neural network model compression methods can effectively optimize the storage space and inference time of the model while maintaining the accuracy. Although automated pruning methods can avoid designing pruning rules, they face the problem of search efficiency when optimizing complex networks. In this paper, a network collaborative pruning method is proposed for hyperspectral image classification based on evolutionary multi-task optimization. The proposed method allows classification networks to perform the model pruning task on multiple hyperspectral images simultaneously. Knowledge (the important local sparse structure of the network) is automatically searched and updated by using knowledge transfer between different tasks. The self-adaptive knowledge transfer strategy based on historical information and dormancy mechanism is designed to avoid possible negative transfer and unnecessary consumption of computing resources. The pruned networks can achieve high classification accuracy on hyperspectral data with limited labeled samples. Experiments on multiple hyperspectral images show that the proposed method can effectively realize the compression of the network model and the classification of hyperspectral images.