Unsupervised Information Research Articles

Spatial-Temporal Dynamic Hypergraph Information Bottleneck for Brain Network Classification.

Recently, Graph Neural Networks (GNNs) have gained widespread application in automatic brain network classification tasks, owing to their ability to directly capture crucial information in non-Euclidean structures. However, two primary challenges persist in this domain. First, within the realm of clinical neuro-medicine, signals from cerebral regions are inevitably contaminated with noise stemming from physiological or external factors. The construction of brain networks heavily relies on set thresholds and feature information within brain regions, making it susceptible to the incorporation of such noises into the brain topology. Additionally, the static nature of the artificially constructed brain network's adjacent structure restricts real-time changes in brain topology. Second, mainstream GNN-based approaches tend to focus solely on capturing information interactions of nearest neighbor nodes, overlooking high-order topology features. In response to these challenges, we propose an adaptive unsupervised Spatial-Temporal Dynamic Hypergraph Information Bottleneck (ST-DHIB) framework for dynamically optimizing brain networks. Specifically, adopting an information theory perspective, Graph Information Bottleneck (GIB) is employed for purifying graph structure, and dynamically updating the processed input brain signals. From a graph theory standpoint, we utilize the designed Hypergraph Neural Network (HGNN) and Bi-LSTM to capture higher-order spatial-temporal context associations among brain channels. Comprehensive patient-specific and cross-patient experiments have been conducted on two available datasets. The results demonstrate the advancement and generalization of the proposed framework.

Few-shot Learning for Heterogeneous Information Networks

Heterogeneous information networks (HINs) are a key resource in many domain-specific retrieval and recommendation scenarios and in conversational environments. Current approaches to mining graph data often rely on abundant supervised information. However, supervised signals for graph learning tend to be scarce for a new task and only a handful of labeled nodes may be available. Meta-learning mechanisms are able to harness prior knowledge that can be adapted to new tasks. In this article, we design meta-learning framework for heterogeneous information networks ( META-HIN ), for few-shot learning problems on HINs. To the best of our knowledge, we are among the first to design a unified framework to realize the few-shot learning of HINs and facilitate different downstream tasks across different domains of graphs. Unlike most previous models, which focus on a single task on a single graph, META-HIN is able to deal with different tasks (node classification, link prediction, and anomaly detection are used as examples) across multiple graphs. Subgraphs are sampled to build the support and query set. Before being processed by the meta-learning module, subgraphs are modeled via a structure module to capture structural features. Then, a heterogeneous Graph Neural Network module is used as the base model to express the features of subgraphs. We also design a Generative Adversarial Network-based contrastive learning module that is able to exploit unsupervised information of the subgraphs. In our experiments, we fuse several datasets from multiple domains to verify META-HIN ’s broad applicability in a multiple-graph scenario. META-HIN consistently and significantly outperforms state-of-the-art alternatives on every task and across all datasets that we consider.

Online continual learning through unsupervised mutual information maximization

Catastrophic forgetting remains a challenge for artificial learning systems, especially in the case of Online learning, where task information is unavailable. This work proposes a novel task-agnostic approach for class-incremental Continual Learning that combines expansion and regularization methods through a non-parametric model. The model adds new experts automatically while regularizing old experts under a variational Bayes paradigm. Expert selection is implemented using Contrastive Learning techniques, catastrophic forgetting is mitigated by minimizing mutual information between the experts’ posterior and prior feature embeddings. Importantly, our method successfully handles single source tasks such as split-MNIST and split-CIFAR-10/100, mixed source tasks like split-CIFAR-100/split-F-CelebA, and long task sequences such as split-TinyImageNet, without using generative models or replay mechanisms. We also successfully extend our approach to Continual Reinforcement learning tasks. Our method achieves these results in a task-agnostic and replay-free setting, making it more flexible than most existing Continual Learning approaches without compromising performance. We provide the code and hyper-parameters for our experiments.

Sd-net: a semi-supervised double-cooperative network for liver segmentation from computed tomography (CT) images

IntroductionThe automatic segmentation of the liver is a crucial step in obtaining quantitative biomarkers for accurate clinical diagnosis and computer-aided decision support systems. This task is challenging due to the frequent presence of noise and sampling artifacts in computerized tomography (CT) images, as well as the complex background, variable shapes, and blurry boundaries of the liver. Standard segmentation of medical images based on full-supervised convolutional networks demands accurate dense annotations. Such a learning framework is built on laborious manual annotation with strict requirements for expertise, leading to insufficient high-quality labels. MethodsTo overcome such limitation and exploit massive weakly labeled data, we relaxed the rigid labeling requirement and developed a semi-supervised double-cooperative network (SD- Net). SD-Net is trained to segment the complete liver volume from preoperative abdominal CT images by using limited labeled datasets and large-scale unlabeled datasets. Specifically, to enrich the diversity of unsupervised information, we construct SD-Net consisting of two collaborative network models. Within the supervised training module, we introduce an adaptive mask refinement approach. First, each of the two network models predicts the labeled dataset, after which adaptive mask refinement of the difference predictions is implemented to obtain more accurate liver segmentation results. In the unsupervised training module, a dynamic pseudo-label generation strategy is proposed. First each of the two models predicts unlabeled data and the better prediction is considered as pseudo-labeling before training. Results and discussionBased on the experimental findings, the proposed method achieves a dice score exceeding 94%, indicating its high level of accuracy and its suitability for everyday clinical use.

Discriminative Regression With Adaptive Graph Diffusion.

In this article, we propose a new linear regression (LR)-based multiclass classification method, called discriminative regression with adaptive graph diffusion (DRAGD). Different from existing graph embedding-based LR methods, DRAGD introduces a new graph learning and embedding term, which explores the high-order structure information between four tuples, rather than conventional sample pairs to learn an intrinsic graph. Moreover, DRAGD provides a new way to simultaneously capture the local geometric structure and representation structure of data in one term. To enhance the discriminability of the transformation matrix, a retargeted learning approach is introduced. As a result of combining the above-mentioned techniques, DRAGD can flexibly explore more unsupervised information underlying the data and the label information to obtain the most discriminative transformation matrix for multiclass classification tasks. Experimental results on six well-known real-world databases and a synthetic database demonstrate that DRAGD is superior to the state-of-the-art LR methods.

A Dual-channel Semi-supervised Learning Framework on Graphs via Knowledge Transfer and Meta-learning

This article studies the problem of semi-supervised learning on graphs, which aims to incorporate ubiquitous unlabeled knowledge (e.g., graph topology, node attributes) with few-available labeled knowledge (e.g., node class) to alleviate the scarcity issue of supervised information on node classification. While promising results are achieved, existing works for this problem usually suffer from the poor balance of generalization and fitting ability due to the heavy reliance on labels or task-agnostic unsupervised information. To address the challenge, we propose a dual-channel framework for semi-supervised learning on G raphs via K nowledge T ransfer between independent supervised and unsupervised embedding spaces, namely, GKT. Specifically, we devise a dual-channel framework including a supervised model for learning the label probability of nodes and an unsupervised model for extracting information from massive unlabeled graph data. A knowledge transfer head is proposed to bridge the gap between the generalization and fitting capability of the two models. We use the unsupervised information to reconstruct batch-graphs to smooth the label probability distribution on the graphs to improve the generalization of prediction. We also adaptively adjust the reconstructed graphs by encouraging the label-related connections to solidify the fitting ability. Since the optimization of the supervised channel with knowledge transfer contains that of the unsupervised channel as a constraint and vice versa, we then propose a meta-learning-based method to solve the bi-level optimization problem, which avoids the negative transfer and further improves the model’s performance. Finally, extensive experiments validate the effectiveness of our proposed framework by comparing state-of-the-art algorithms.

Enhancing Hyperspectral Image Classification: Leveraging Unsupervised Information With Guided Group Contrastive Learning

Enhancing Hyperspectral Image Classification: Leveraging Unsupervised Information With Guided Group Contrastive Learning

Latent Linear Discriminant Analysis for feature extraction via Isometric Structural Learning

Linear discriminant analysis (LDA) is one of the most successful feature extraction methods, which projects high-dimensional data to a low-dimensional space with discriminative features. However, there are problems in the existing LDAs: (1) the effect of hidden data is not exploited in LDA, (2) the LDAs cannot preserve the local isometric structure, (3) there is no consideration for structural consistency that unifies the supervised global and unsupervised local information. In this paper, we propose a brand-new LDA method, namely, Latent Linear Discriminant Analysis with Isometric Structural Learning (L2DA-ISL). We formulate LDA to a latent representation framework that considers both the discriminability from observed data and hidden data. Then, we propose isometric structural learning to capture the intrinsic local structural information. Lastly, we establish the concept of structural consistency in LDA framework. Extensive experiments and comparisons show that L2DA-ISL achieves a promising performance with structural consistency and stronger robustness in feature extraction.

A data expansion technique based on training and testing sample to boost the detection of SSVEPs for brain-computer interfaces

Objective. Currently, steady-state visual evoked potentials (SSVEPs)-based brain-computer interfaces (BCIs) have achieved the highest interaction accuracy and speed among all BCI paradigms. However, its decoding efficacy depends deeply on the number of training samples, and the system performance would have a dramatic drop when the training dataset decreased to a small size. To date, no study has been reported to incorporate the unsupervised learning information from testing trails into the construction of supervised classification model, which is a potential way to mitigate the overfitting effect of limited samples. Approach. This study proposed a novel method for SSVEPs detection, i.e. cyclic shift trials (CSTs), which could combine unsupervised learning information from test trials and supervised learning information from train trials. Furthermore, since SSVEPs are time-locked and phase-locked to the onset of specific flashes, CST could also expand training samples on the basis of its regularity and periodicity. In order to verify the effectiveness of CST, we designed an online SSVEP-BCI system, and tested this system combined CST with two common classification algorithms, i.e. extended canonical correlation analysis and ensemble task-related component analysis. Main results. CST could significantly enhance the signal to noise ratios of SSVEPs and improve the performance of systems especially for the condition of few training samples and short stimulus time. The online information transfer rate could reach up to 236.19 bits min−1 using 36 s calibration time of only one training sample for each category. Significance. The proposed CST method can take full advantages of supervised learning information from training samples and unsupervised learning information of testing samples. Furthermore, it is a data expansion technique, which can enhance the SSVEP characteristics and reduce dependence on sample size. Above all, CST is a promising method to improve the performance of SSVEP-based BCI without any additional experimental burden.

SUSIE: Pharmaceutical CMC ontology-based information extraction for drug development using machine learning

Automatically extracting information from unstructured text in pharmaceutical documents is important for drug discovery and development. This information can be integrated with structured datasets to ultimately accelerate pharmaceutical product development. To this end, we report an end-to-end information extraction framework based on a custom-built pharmaceutical drug development ontology, a weak supervision framework, contextualization algorithms, and a fine-tuned BioBERT model (adaptation of BERT or Bidirectional Encoder Representations from Transformers for biomedical text). The proposed framework, SUSIE (Schema-based Unsupervised Semantic Information Extraction), was trained on ICH (International Conference on Harmonization) documents to identify important entities and relations from unstructured text and auto-generate knowledge graphs representing crucial information in a structured format. On the entity identification task, the framework achieves a test accuracy and F1-score of 96% and 88%, respectively, on out-of-sample documents. A major contribution of this work is to build an automated, unsupervised information extraction framework around a domain-specific, custom-built pharmaceutical drug development ontology without the need for manual curation of training datasets for specific tasks. The efficacy of the approach was tested on out-of-sample documents including an internal Eli Lilly technical document.

Extraction, labelling, clustering, and semantic mapping of segments from clinical notes.

This work is motivated by the scarcity of tools for accurate, unsupervised information extraction from unstructured clinical notes in computationally underrepresented languages, such as Czech. We introduce a stepping stone to a broad array of downstream tasks such as summarisation or integration of individual patient records, extraction of structured information for national cancer registry reporting or building of semi-structured semantic patient representations that can be used for computing patient embeddings. More specifically, we present a method for unsupervised extraction of semantically-labelled textual segments from clinical notes and test it out on a dataset of Czech breast cancer patients, provided by Masaryk Memorial Cancer Institute (the largest Czech hospital specialising exclusively in oncology). Our goal was to extract, classify (i.e. label) and cluster segments of the free-text notes that correspond to specific clinical features (e.g., family background, comorbidities or toxicities). Finally, we propose a tool for computer-assisted semantic mapping of segment types to pre-defined ontologies and validate it on a downstream task of category-specific patient similarity. The presented results demonstrate the practical relevance of the proposed approach for building more sophisticated extraction and analytical pipelines deployed on Czech clinical notes.

A detailed library perspective on nearly unsupervised information extraction workflows in digital libraries

Information extraction can support novel and effective access paths for digital libraries. Nevertheless, designing reliable extraction workflows can be cost-intensive in practice. On the one hand, suitable extraction methods rely on domain-specific training data. On the other hand, unsupervised and open extraction methods usually produce not-canonicalized extraction results. This paper is an extension of our original work and tackles the question of how digital libraries can handle such extractions and whether their quality is sufficient in practice. We focus on unsupervised extraction workflows by analyzing them in case studies in the domains of encyclopedias (Wikipedia), Pharmacy, and Political Sciences. As an extension, we analyze the extractions in more detail, verify our findings on a second extraction method, discuss another canonicalizing method, and give an outlook on how non-English texts can be handled. Therefore, we report on opportunities and limitations. Finally, we discuss best practices for unsupervised extraction workflows.

An Intelligent Conversational Agent for the Legal Domain

An intelligent conversational agent for the legal domain is an AI-powered system that can communicate with users in natural language and provide legal advice or assistance. In this paper, we present CREA2, an agent designed to process legal concepts and be able to guide users on legal matters. The conversational agent can help users navigate legal procedures, understand legal jargon, and provide recommendations for legal action. The agent can also give suggestions helpful in drafting legal documents, such as contracts, leases, and notices. Additionally, conversational agents can help reduce the workload of legal professionals by handling routine legal tasks. CREA2, in particular, will guide the user in resolving disputes between people residing within the European Union, proposing solutions in controversies between two or more people who are contending over assets in a divorce, an inheritance, or the division of a company. The conversational agent can later be accessed through various channels, including messaging platforms, websites, and mobile applications. This paper presents a retrieval system that evaluates the similarity between a user’s query and a given question. The system uses natural language processing (NLP) algorithms to interpret user input and associate responses by addressing the problem as a semantic search similar question retrieval. Although a common approach to question and answer (Q&A) retrieval is to create labelled Q&A pairs for training, we exploit an unsupervised information retrieval system in order to evaluate the similarity degree between a given query and a set of questions contained in the knowledge base. We used the recently proposed SBERT model for the evaluation of relevance. In the paper, we illustrate the effective design principles, the implemented details and the results of the conversational system and describe the experimental campaign carried out on it.

Boundary-enhanced semi-supervised retinal layer segmentation in optical coherence tomography images using fewer labels.

Automatic segmentation of multiple layers in retinal optical coherence tomography (OCT) images is crucial for eye disease diagnosis and treatment. Despite the success of deep learning algorithms, it still remains a challenge due to the blurry layer boundaries and lack of adequate pixel-wise annotations. To tackle these issues, we propose a Boundary-Enhanced Semi-supervised Network (BE-SemiNet) that exploits an auxiliary distance map regression task to improve retinal layer segmentation with scarce labeled data and abundant unlabeled data. Specifically, a novel Unilaterally Truncated Distance Map (UTDM) is firstly introduced to alleviate the class imbalance problem and enhance the layer boundary learning in the regression task. Then for the pixel-wise segmentation and UTDM regression branches, we impose task-level and data-level consistency regularization on unlabeled data to enrich the diversity of unsupervised information and improve the regularization effects. Pseudo supervision is incorporated in consistency regularization to bridge the task prediction spaces for consistency and expand training labeled data. Experiments on two public retinal OCT datasets show that our method can greatly improve the supervised baseline performance with only 5 annotations and outperform the state-of-the-art methods. Since it is difficult and labor-expensive to obtain adequate pixel-wise annotations in practice, our method has a promising application future in clinical retinal OCT image analysis.

Housing rental suggestion based on e-commerce data

Because renting a house is a low-frequency behavior of people, the housing rental suggestion issue suffers from extremely sparse data. In this paper, we propose to investigate the issue based on e-commerce data, as the two scenarios share the same users, and, more importantly, we can characterize the users in terms of their consumption attitudes and behaviors (that can direct people to rent houses) from the massive online shopping orders, and extract labeled users (users who have both housing and working addresses) from the delivery records. Though e-commerce data can be utilized, developing such an approach is still challenging. This happens because only a tiny part of these users can be used as supervised information and it is non-trivial to fuse the e-commerce data with geographic and traffic data. To this end, we first propose to formulate the task pairwise way to use the labeled users more effectively, the target of which is to evaluate a user’s satisfaction degree for a given house. Thereafter, we carefully design features of users and houses from the multi-source data and devise a novel deep network, entitled HouseCritic, to fuse the features and tackle the formulated problem. Notably, HouseCritic introduces meta-learning technology to design a fusion structure that can explicitly and effectively fuse extracted features and accordingly infer a satisfaction degree. Moreover, HouseCritic proposes to leverage unsupervised information, capable of effectively fusing data when the supervised information is insufficient. Finally, we conduct experiments on real-world data collected from Beijing, China. Experimental results demonstrate the advantages of the proposed approach over several baselines. Moreover, a system with the approach is being deployed in an e-commerce company.

Hyperspectral Image Classification Based on Unsupervised Regularization

Due to the powerful feature expression ability of deep learning and its end-to-end non-linear mapping relationship, deep learning-based methods have become the mainstream method for hyperspectral image classification tasks. However, the accuracy of deep learning methods greatly depends on the use of a large number of labeled samples to train the model. Also, hyperspectral images have few labeled samples and unbalanced categories, which makes the depth model prone to overfitting, which seriously affects the classification accuracy. Therefore, how to alleviate the overfitting phenomenon caused by small samples in the classification problem based on deep learning is still a problem that needs to be solved. Considering that it is relatively easier to obtain a large number of unlabeled samples in the field of remote sensing, making full use of the unsupervised information learned from unlabeled data can regularize the supervised classification model, which can effectively alleviate the overfitting phenomenon caused by the small samples problem. In the supervised training process, unsupervised information from the overall distribution of the sample is introduced to guide the regularization of the model, so as to realize the effective classification of the data in the case of a small number of labeled samples. Experimental results demonstrate the effectiveness of the proposed method in terms of hyperspectral image (HSI) classification with few training samples.

Deep learning approach for segmentation and classification of blood cells using enhanced CNN

The main aim of our proposed work is to segment and classify blood cells using K means algorithm, and also with the help of image processing techniques. A complete blood cell count is very important in the medical analysis for evaluating the total health condition of the body. In the olden days, the blood cells are counted manually with the help of a hem cytometer with other lab equipment and certain chemicals. This technique both time-consuming and challenging. Deep Learning (DL) is an artificial intelligence subset of machine learning that can examine unsupervised information. RBC, or red blood cells, are red cells that are numerous in the blood and are noted for their dark red colour. White blood cells, commonly known as leukocytes, protect the body from infection. Processing tools like MATLAB is used to find the variations in area, perimeter and statistical parameters like mean and standard deviation that separates white blood cells from other blood components. Because of its high accuracy, Enhanced CNN can be used in this study for the categorization and recognition of normal and abnormal blood cell images. Average accuracy of 95% and average precision 0f 0.93 which is higher than existing CNN.

AFSNet: attention-guided full-scale feature aggregation network for high-resolution remote sensing image change detection

ABSTRACT Using change detection technique precisely analyzes remote sensing images, it has a broad range of applications in resource surveys, surveillance systems, and map updating. In recent years, deep learning-based methods have become a focus area owing to their excellent feature extraction and representation ability. The fusion of multi-scale features is the key to improving change detection performance in fully convolutional network-based structural methods, primarily based on an architecture with skip connections or nested and dense skip connections. However, these methods only fuse features on the same scale and lack sufficient information from multiple scales to generate appropriate results. Traditional feature fusion with redundant and unsupervised information leads to poor model fitting. To solve these problems, we proposed a novel attention-guided full-scale feature aggregation network (AFSNet). The proposed method used a Siamese structure as the backbone network to extract features, which were then aggregated using full-scale skip connections, and an attention mechanism to avoid feature redundancy. Finally, to obtain a highly accurate final change map, a multiple side-outputs fusion strategy was used to fuse the change maps at different scales. To check the reliability of AFSNet, we tested it on two public datasets, the LEVIR-CD and SVCD datasets. The F1-Score/IoU scores improved by 0.57%/0.95% and 1.14%/2.07% in the two datasets, respectively, compared to those obtained using the methods that achieved suboptimal values. The results showed that AFSNet outperforms other mainstream methods while maintaining a good balance between computational costs and model parameters.

Multi-rater label fusion based on an information bottleneck for fundus image segmentation

In the fundus image segmentation process, more than one professional rater is usually required to mark the organ in the image to reduce the error rate of diagnosis. However, it is undeniable that each rater has an independent level of expertise and experience, and there are often individual differences in annotation areas, which can produce redundant or irrelevant information. In computer vision, for feature selection of multiple annotations, the majority voting principle or the preferred annotator principle is usually used, which cannot effectively remove the noise inside the annotator and the redundant information between annotators. To solve the above problems, we studied the information bottleneck method to remove the annotation noise and extend it to multi-expert annotation tasks in the fundus image field to extract consistent information between different views. We believe that consistent information is the key information that multiple experts agree on and have directivity. To the best of our knowledge, this is the first model for extracting multi-expert consistency information via multi-view information bottlenecks. Specifically, we use a multi-view information bottleneck approach to obtain the most concise expression of each view under label supervision. In addition, we propose a novel unsupervised information bottleneck method by maximizing mutual information between multiple view representations to preserve consistent information while eliminating redundant information that is not shared between views. A large number of experiments on several public datasets prove the effectiveness of the proposed model, and its performance is superior to existing techniques.

IGNFusion: An Unsupervised Information Gate Network for Multimodal Medical Image Fusion

Multimodal medical image fusion aims to merge saliency and complementary information from different source images to assist in biomedical diagnoses. How to effectively utilize feature information in the encoder is a critical issue. However, many existing medical image fusion methods do not consider the contributions of different convolution blocks. In this paper, we propose an information gate module (IGM) to control the contribution of each encoder feature level to the decoder; it is termed the information gate network for multimodal medical image fusion (IGNFusion). Furthermore, the Siamese multi-scale cross attention fusion module (SMSCAFM) integrates saliency and complementary information from multiple source images. Moreover, to constrain the similarity between the fused image and multiple source images, we introduce a saliency weight (SW). Extensive experiments on ten categories of multimodal medical images (i.e., CT <inline-formula><tex-math notation="LaTeX">$\&amp; $</tex-math></inline-formula> MR-T1 (T1 weighted) and PET <inline-formula><tex-math notation="LaTeX">$\&amp; $</tex-math></inline-formula> MR-T2 (T2 weighted)) show that our IGNFusion approach achieves significant improvements over 9 state-of-the-art methods.