Article
Computer Science, Artificial Intelligence
Handong Ma, Changsheng Li, Xinchu Shi, Ye Yuan, Guoren Wang
Summary: This paper introduces a new deep unsupervised active learning model that utilizes learnable graphs to improve sample representation and selection of representative samples. By learning optimal graph structures and incorporating shortcut connections, this approach achieves good results in unsupervised active learning.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Interdisciplinary Applications
Ilkay Yildiz, Rachael Garner, Matthew Lai, Dominique Duncan
Summary: This article introduces an unsupervised seizure identification method based on deep learning. The method uses a variational autoencoder (VAE) to train on raw EEG and identifies seizures based on reconstruction errors. The experimental results show that our method can successfully distinguish seizures from non-seizure activity.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Agriculture, Multidisciplinary
Hongfei Zhu, Xingyu Liu, Hao Zheng, Lianhe Yang, Xuchen Li, Zhongzhi Han
Summary: In this study, a new strawberry appearance quality detection method based on unsupervised deep learning was proposed. It extracted strawberry image features using deep learning models (Resnet18, Resnet50, and Resnet101), reduced the dimension of feature vectors using t-SNE, and performed cluster analysis using Gaussian Mixture Model. The results showed high accuracy in clustering strawberries based on different deep learning models in 2D and 3D spaces, highlighting the potential of this method for improving intelligent picking of strawberries.
PRECISION AGRICULTURE
(2023)
Article
Geochemistry & Geophysics
Ying Qu, Razieh Kaviani Baghbaderani, Hairong Qi, Chiman Kwan
Summary: The article proposes an unsupervised pansharpening method based on a deep learning framework to address the challenge of fully utilizing the rich spectral characteristics in MSI. The method, based on the self-attention mechanism, can extract and inject details with subpixel accuracy, and experimental results demonstrate its effectiveness in reconstructing sharper MSI.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Computer Science, Artificial Intelligence
Hui Xu, Jiaxing Wang, Hao Li, Deqiang Ouyang, Jie Shao
Summary: This paper proposes an unsupervised meta-learning algorithm that learns from an unlabeled dataset and adapts to downstream human specific tasks with few labeled data. Experimental results show that the proposed method outperforms other tested unsupervised representation learning approaches and two recent unsupervised meta-learning baselines on two datasets.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Artificial Intelligence
Marcus de Carvalho, Mahardhika Pratama, Jie Zhang, Edward Yapp Kien Yee
Summary: This article discusses the problem of online unsupervised cross-domain adaptation and proposes the ACDC framework. It effectively addresses the challenges in data streams through a self-evolving neural network structure and achieves promising results in experiments.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Han Zhao, Xu Yang, Cheng Deng, Dacheng Tao
Summary: In this study, we propose a structure-adaptive graph contrastive learning framework to capture potential discriminative relationships for improved graph representation learning.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Chengzhang Zhu, Longbing Cao, Jianping Yin
Summary: This paper introduces a shallow but powerful unsupervised learning method called UNTIE for representing coupled categorical data. It reveals heterogeneous distributions between couplings and achieves significant performance improvement on multiple categorical datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Mathematics, Applied
Yibin Wang, Haifeng Wang
Summary: In this study, a distributionally robust unsupervised domain adaptation (DRUDA) method is proposed to enhance the generalization ability of machine learning models under input space perturbations. The DRUDA approach optimizes worst-case perturbations of the training source data to reduce the shifts in joint distributions across domains, leading to improved domain adaptation accuracy on target domains.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2023)
Article
Materials Science, Multidisciplinary
Yuan-Heng Tseng, Fu-Jiun Jiang
Summary: This study investigates the impact of using various training sets on the performance of an unsupervised neural network (NN) for learning the phases of a two-dimensional ferromagnetic Potts model, specifically a deep learning autoencoder (AE). The results show that data below and near the transition temperature T-c are crucial in successfully training the AE. Additionally, the commonly used training procedures for unsupervised NNs are found to be inefficient, and the findings from this study can serve as useful guidelines for setting up effective trainings for unsupervised NNs.
RESULTS IN PHYSICS
(2022)
Article
Computer Science, Artificial Intelligence
Jiang Lu, Lei Li, Changshui Zhang
Summary: The remarkable progress in deep learning largely relies on large-scale supervised data. Ensuring intra-class modality diversity in the training set is crucial for the generalization capability of cutting-edge deep models, but it requires heavy manual labor for data collection and annotation. Additionally, rare or unexpected modalities may cause reduced performance in current models under emerging modalities.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Materials Science, Multidisciplinary
Courtney Kunselman, Sofia Sheikh, Madalyn Mikkelsen, Vahid Attari, Raymundo Arroyave
Summary: The framework developed in this work reduces the cost of human annotation by leveraging novel machine learning procedures for class discovery and label assignment. It utilizes semi-supervised classification to combine high-and low-confidence label assignments, resulting in highly accurate classifiers for microstructure image class taxonomies discovered solely through data-driven methods.
Article
Geochemistry & Geophysics
Jian Kang, Ruben Fernandez-Beltran, Puhong Duan, Sicong Liu, Antonio J. Plaza
Summary: The article presents a new unsupervised deep metric learning model called SauMoCo, designed to characterize unlabeled RS scenes by defining spatial augmentation criteria and constructing a queue of deep embeddings. The proposed approach substantially enhances the discrimination ability among complex land cover categories of RS tiles.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Computer Science, Artificial Intelligence
Chunyang Cheng, Tianyang Xu, Xiao-Jun Wu
Summary: Existing image fusion approaches using a single network often yield suboptimal results due to the lack of ground-truth output. We propose a self-evolutionary training formula with a novel memory unit architecture (MUFusion) that utilizes intermediate fusion results for collaborative supervision. An adaptive unified loss function based on the memory unit is designed to improve fusion quality. Our MUFusion achieves superior performance in various image fusion tasks according to qualitative and quantitative experiments.
INFORMATION FUSION
(2023)
Article
Engineering, Aerospace
Mengqiu Xu, Ming Wu, Jun Guo, Chuang Zhang, Yubo Wang, Zhanyu Ma
Summary: Sea fog detection with remote sensing images is a challenging task due to the lack of meteorological observations and buoys over the sea for obtaining visibility information. This article proposes an unsupervised domain adaptation method to bridge labeled land fog data and unlabeled sea fog data, achieving sea fog detection by leveraging the similarity between land fog and sea fog.
CHINESE JOURNAL OF AERONAUTICS
(2022)