A comparison review of transfer learning and self-supervised learning: Definitions, applications, advantages and limitations

Expert Systems with Applications, Год журнала: 2023, Номер 242, С. 122807 - 122807

Опубликована: Дек. 2, 2023

Deep learning has emerged as a powerful tool in various domains, revolutionising machine research. However, one persistent challenge is the scarcity of labelled training data, which hampers performance and generalisation deep models. To address this limitation, researchers have developed innovative methods to overcome data enhance model capabilities. Two prevalent techniques that gained significant attention are transfer self-supervised learning. Transfer leverages knowledge learned from pre-training on large-scale dataset, such ImageNet, applies it target task with limited data. This approach allows models benefit representations effectively new tasks, resulting improved generalisation. On other hand, focuses using pretext tasks do not require manual annotation, allowing them learn valuable large amounts unlabelled These can then be fine-tuned for downstream mitigating need extensive In recent years, found applications fields, including medical image processing, video recognition, natural language processing. approaches demonstrated remarkable achievements, enabling breakthroughs areas disease diagnosis, object understanding. while these offer numerous advantages, they also limitations. For example, may face domain mismatch issues between requires careful design ensure meaningful representations. review paper explores fields within past three years. It delves into advantages limitations each approach, assesses employing techniques, identifies potential directions future By providing comprehensive current methods, article offers guidance selecting best technique specific issue.

Язык: Английский

---

Comprehensive systematic review of information fusion methods in smart cities and urban environments

Information Fusion, Год журнала: 2024, Номер 107, С. 102317 - 102317

Опубликована: Фев. 21, 2024

Smart cities result from integrating advanced technologies and intelligent sensors into modern urban infrastructure. The Internet of Things (IoT) data integration are pivotal in creating interconnected spaces. In this literature review, we explore the different methods information fusion used smart cities, along with their advantages challenges. However, there notable challenges managing diverse sources, handling large volumes, meeting near-real-time demands various city applications. review aims to examine applications detail, incorporating quality evaluation techniques identifying critical issues while outlining promising research directions. order accomplish our goal, conducted a comprehensive search applied selective criteria. We identified 59 recent studies addressing machine learning (ML) deep (DL) These were obtained databases such as ScienceDirect (SD), Scopus, Web Science (WoS), IEEE Xplore. main objective study is provide more detailed insights by supplementing existing research. word cloud visualisation learning/deep papers shows landscape, covering both technical aspects artificial intelligence practical settings. Apart exploration, also delves ethical privacy implications arising cities. Moreover, it thoroughly examines that must be addressed realise revolution's potential fully.

Язык: Английский

---

Explainable AI-driven IoMT fusion: Unravelling techniques, opportunities, and challenges with Explainable AI in healthcare

Information Fusion, Год журнала: 2024, Номер 110, С. 102472 - 102472

Опубликована: Май 16, 2024

Язык: Английский

---

Trustworthy deep learning framework for the detection of abnormalities in X-ray shoulder images

PLoS ONE, Год журнала: 2024, Номер 19(3), С. e0299545 - e0299545

Опубликована: Март 11, 2024

Musculoskeletal conditions affect an estimated 1.7 billion people worldwide, causing intense pain and disability. These lead to 30 million emergency room visits yearly, the numbers are only increasing. However, diagnosing musculoskeletal issues can be challenging, especially in emergencies where quick decisions necessary. Deep learning (DL) has shown promise various medical applications. previous methods had poor performance a lack of transparency detecting shoulder abnormalities on X-ray images due training data better representation features. This often resulted overfitting, generalisation, potential bias decision-making. To address these issues, new trustworthy DL framework been proposed detect (such as fractures, deformities, arthritis) using images. The consists two parts: same-domain transfer (TL) mitigate imageNet mismatch feature fusion reduce error rates improve trust final result. Same-domain TL involves pre-trained models large number labelled from body parts fine-tuning them target dataset Feature combines extracted features with seven train several ML classifiers. achieved excellent accuracy rate 99.2%, F1 Score Cohen’s kappa 98.5%. Furthermore, results was validated three visualisation tools, including gradient-based class activation heat map (Grad CAM), visualisation, locally interpretable model-independent explanations (LIME). outperformed orthopaedic surgeons invited classify test set, who obtained average 79.1%. proven effective robust, improving generalisation increasing results.

Язык: Английский

---

ATD Learning: A secure, smart, and decentralised learning method for big data environments

Information Fusion, Год журнала: 2025, Номер unknown, С. 102953 - 102953

Опубликована: Янв. 1, 2025

Язык: Английский

---

Shedding light on ai in radiology: A systematic review and taxonomy of eye gaze-driven interpretability in deep learning

European Journal of Radiology, Год журнала: 2024, Номер 172, С. 111341 - 111341

Опубликована: Фев. 2, 2024

X-ray imaging plays a crucial role in diagnostic medicine. Yet, significant portion of the global population lacks access to this essential technology due shortage trained radiologists. Eye-tracking data and deep learning models can enhance analysis by mapping expert focus areas, guiding automated anomaly detection, optimizing workflow efficiency, bolstering training methods for novice However, literature shows contradictory results regarding usefulness eye-tracking deep-learning architectures abnormality detection. We argue that these discrepancies between studies are (a) way is (or not) processed, (b) types chosen, (c) type application will have. conducted systematic review using PRISMA address contradicting results. analyzed 60 incorporated approach different goals radiology. performed comparative understand if eye gaze contains feature maps be useful under whether they promote more interpretable predictions. To best our knowledge, first survey area performs thorough investigation processing techniques their impacts applications such as error classification, object expertise level analysis, fatigue estimation human attention prediction medical data. Our resulted two main contributions: (1) taxonomy divides task, enabling us analyze value movement bring each case build guidelines adequate application, (2) an overall how explainability

Язык: Английский

---

The METRIC-framework for assessing data quality for trustworthy AI in medicine: a systematic review

npj Digital Medicine, Год журнала: 2024, Номер 7(1)

Опубликована: Авг. 3, 2024

The adoption of machine learning (ML) and, more specifically, deep (DL) applications into all major areas our lives is underway. development trustworthy AI especially important in medicine due to the large implications for patients' lives. While trustworthiness concerns various aspects including ethical, transparency and safety requirements, we focus on importance data quality (training/test) DL. Since dictates behaviour ML products, evaluating will play a key part regulatory approval medical products. We perform systematic review following PRISMA guidelines using databases Web Science, PubMed ACM Digital Library. identify 5408 studies, out which 120 records fulfil eligibility criteria. From this literature, synthesise existing knowledge frameworks combine it with perspective medicine. As result, propose METRIC-framework, specialised framework training comprising 15 awareness dimensions, along developers should investigate content dataset. This helps reduce biases as source unfairness, increase robustness, facilitate interpretability thus lays foundation METRIC-framework may serve base systematically assessing datasets, establishing reference designing test datasets has potential accelerate

Язык: Английский

---

Backbones-review: Feature extractor networks for deep learning and deep reinforcement learning approaches in computer vision

Computer Science Review, Год журнала: 2024, Номер 53, С. 100645 - 100645

Опубликована: Июнь 7, 2024

Язык: Английский

---

Comprehensive review of deep learning in orthopaedics: Applications, challenges, trustworthiness, and fusion

Artificial Intelligence in Medicine, Год журнала: 2024, Номер 155, С. 102935 - 102935

Опубликована: Июль 26, 2024

Deep learning (DL) in orthopaedics has gained significant attention recent years. Previous studies have shown that DL can be applied to a wide variety of orthopaedic tasks, including fracture detection, bone tumour diagnosis, implant recognition, and evaluation osteoarthritis severity. The utilisation is expected increase, owing its ability present accurate diagnoses more efficiently than traditional methods many scenarios. This reduces the time cost diagnosis for patients surgeons. To our knowledge, no exclusive study comprehensively reviewed all aspects currently used practice. review addresses this knowledge gap using articles from Science Direct, Scopus, IEEE Xplore, Web between 2017 2023. authors begin with motivation orthopaedics, enhance treatment planning. then covers various applications detection supraspinatus tears MRI, osteoarthritis, prediction types arthroplasty implants, age assessment, joint-specific soft tissue disease. We also examine challenges implementing scarcity data train lack interpretability, as well possible solutions these common pitfalls. Our work highlights requirements achieve trustworthiness outcomes generated by DL, need accuracy, explainability, fairness models. pay particular fusion techniques one ways increase trustworthiness, which been address multimodality orthopaedics. Finally, we approval set forth US Food Drug Administration enable use applications. As such, aim function guide researchers develop reliable application tasks scratch market.

Язык: Английский

---

Is Artificial Intelligence (AI) Research Biased and Conceptually Vague? A Systematic Review of Research on Bias and Discrimination in the context of using AI in Human Resource Management

Technology in Society, Год журнала: 2025, Номер unknown, С. 102818 - 102818

Опубликована: Янв. 1, 2025

Язык: Английский

---