Lina Ge
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lina Ge has a total of 74 co-authors. Here are the top co-authors by number of shared publications:
Guifen Zhang
School of artificial intelligence, GUANGXI MINZU UNIVERSITY, Nanning, China
6 shared publicationsZhe Wang
4 shared publications- Feng Han
School of Artificial Intelligence, Guangxi University for Nationalities, No. 188 Daxue East Road, Nanning 530006, China
3 shared publications - Guang-Hui Dong
Sun Yat-sen University
3 shared publications Lei Tian
3 shared publications- Lei Tian
School of Artificial Intelligence, Guangxi Minzu University, Nanning, Guangxi, China
3 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1314-1324
Continual Knowledge Graph Embedding via Community Hierarchies and Reconstruction Distillation
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1314-1324
Open AccessContinual Knowledge Graph Embedding via Community Hierarchies and Reconstruction Distillation
- 18 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
HFedM2ND: A secure Hierarchical Federated Learning framework against poisoning attacks with lightweight multi-layer malicious-node detection
- 18 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
HFedM2ND: A secure Hierarchical Federated Learning framework against poisoning attacks with lightweight multi-layer malicious-node detection
- Open Access
- 12 August 2026
- Published by MDPI AG in Healthcare
- Vol. 14 (16) , 2508
Exploring Co-Occurring Clinical and Treatment Characteristics Associated with In Vitro Fertilization Failure in Polycystic Ovary Syndrome: An Association Rule Mining Approach
Background: Polycystic ovary syndrome (PCOS) is a leading cause of anovulatory infertility and a frequent indication for in vitro fertilization/intracytoplasmic sperm injection (IVF/ICSI). Multivariable regression estimates adjusted associations, whereas association rule mining (ARM) can describe frequently co-occurring attributes; however, ARM outputs are unadjusted and may be sensitive to analytic specification. Objectives: We used ARM alongside multivariable logistic regression and internal robustness analyses to describe combinations of clinical and treatment-related variables that co-occurred with clinical pregnancy failure among women with PCOS undergoing IVF/ICSI. Methods: We retrospectively analyzed 733 deidentified patients with PCOS from the Reproductive Hospital of Guangxi. The original one-hot-encoded analytic file was linked to a supplemental clinical record file containing continuous age, BMI, infertility duration, anti-Müllerian hormone, total gonadotropin dose, fertilization method, embryo-transfer day, number of embryos transferred, and embryo score. Embryo quality was derived from D3 cleavage-stage and blastocyst morphology records. Multivariable logistic regression was used to estimate adjusted associations. ARM results were assessed using Fisher’s exact test, Benjamini–Hochberg false discovery rate (FDR) correction, 1000 bootstrap resamples, fivefold cross-validation, and threshold/cutoff sensitivity analyses. Results: In the multivariable model (n = 730), age remained associated with clinical pregnancy failure (adjusted odds ratio [aOR] = 1.04 per year, 95% CI 1.00–1.08, p = 0.031), whereas BMI was not independently associated with failure (aOR = 1.00 per kg/m2, 95% CI 0.96–1.05, p = 0.860). Transfer of two embryos (aOR = 0.55, 95% CI 0.35–0.85, p = 0.008), D5 blastocyst transfer (aOR = 0.49, 95% CI 0.32–0.75, p = 0.001), and transfer of at least one high-quality embryo (aOR = 0.63, 95% CI 0.43–0.93, p = 0.020) were associated with lower odds of failure. Under the primary host-factor specification, age > 35 years + BMI ≥ 24 kg/m2 + pelvic adhesion defined 48 patients, 30 of whom had clinical pregnancy failure (support = 0.07; confidence = 0.63; lift = 1.38; FDR q = 0.098). In the secondary treatment-inclusive analysis, infertility duration > 5 years + no high-quality embryo transferred defined 43 patients with 30 failures (support = 0.06; confidence = 0.70; lift = 1.54; FDR q = 0.026). Lift represents the subgroup failure proportion divided by the overall failure proportion and is not an adjusted risk ratio. The rule composition and the number of FDR-supported rules changed under alternative cutoffs and confidence thresholds. Conclusions: ARM described transparent but specification-sensitive co-occurrence profiles that complement, rather than compete with, regression. Treatment-inclusive profiles combine baseline and downstream cycle characteristics and are therefore especially exploratory. These findings require prospective multicenter validation before any clinical decision-support use.
- 12 August 2026
- Published by MDPI AG in Healthcare
- Vol. 14 (16) , 2508
Open AccessExploring Co-Occurring Clinical and Treatment Characteristics Associated with In Vitro Fertilization Failure in Polycystic Ovary Syndrome: An Association Rule Mining Approach
Background: Polycystic ovary syndrome (PCOS) is a leading cause of anovulatory infertility and a frequent indication for in vitro fertilization/intracytoplasmic sperm injection (IVF/ICSI). Multivariable regression estimates adjusted associations, whereas association rule mining (ARM) can describe frequently co-occurring attributes; however, ARM outputs are unadjusted and may be sensitive to analytic specification. Objectives: We used ARM alongside multivariable logistic regression and internal robustness analyses to describe combinations of clinical and treatment-related variables that co-occurred with clinical pregnancy failure among women with PCOS undergoing IVF/ICSI. Methods: We retrospectively analyzed 733 deidentified patients with PCOS from the Reproductive Hospital of Guangxi. The original one-hot-encoded analytic file was linked to a supplemental clinical record file containing continuous age, BMI, infertility duration, anti-Müllerian hormone, total gonadotropin dose, fertilization method, embryo-transfer day, number of embryos transferred, and embryo score. Embryo quality was derived from D3 cleavage-stage and blastocyst morphology records. Multivariable logistic regression was used to estimate adjusted associations. ARM results were assessed using Fisher’s exact test, Benjamini–Hochberg false discovery rate (FDR) correction, 1000 bootstrap resamples, fivefold cross-validation, and threshold/cutoff sensitivity analyses. Results: In the multivariable model (n = 730), age remained associated with clinical pregnancy failure (adjusted odds ratio [aOR] = 1.04 per year, 95% CI 1.00–1.08, p = 0.031), whereas BMI was not independently associated with failure (aOR = 1.00 per kg/m2, 95% CI 0.96–1.05, p = 0.860). Transfer of two embryos (aOR = 0.55, 95% CI 0.35–0.85, p = 0.008), D5 blastocyst transfer (aOR = 0.49, 95% CI 0.32–0.75, p = 0.001), and transfer of at least one high-quality embryo (aOR = 0.63, 95% CI 0.43–0.93, p = 0.020) were associated with lower odds of failure. Under the primary host-factor specification, age > 35 years + BMI ≥ 24 kg/m2 + pelvic adhesion defined 48 patients, 30 of whom had clinical pregnancy failure (support = 0.07; confidence = 0.63; lift = 1.38; FDR q = 0.098). In the secondary treatment-inclusive analysis, infertility duration > 5 years + no high-quality embryo transferred defined 43 patients with 30 failures (support = 0.06; confidence = 0.70; lift = 1.54; FDR q = 0.026). Lift represents the subgroup failure proportion divided by the overall failure proportion and is not an adjusted risk ratio. The rule composition and the number of FDR-supported rules changed under alternative cutoffs and confidence thresholds. Conclusions: ARM described transparent but specification-sensitive co-occurrence profiles that complement, rather than compete with, regression. Treatment-inclusive profiles combine baseline and downstream cycle characteristics and are therefore especially exploratory. These findings require prospective multicenter validation before any clinical decision-support use.
- Open Access
- 10 August 2026
- Published by Frontiers Media SA in Frontiers in Blockchain
- Vol. 9, 1882200
Exploring digital skills cultivation through a blockchain-enabled ‘digital-intelligent workshop’
Due to the rapid development of new generation information technology worldwide, universities’ cultivation of digital and intelligent talents has become a necessity of the times. To address the common challenges in traditional practice-oriented teaching, including outdated course design, insufficient authenticity of experimental environments, and cross-institutional trust deficits in student evaluation, this study takes the ‘Digital-Intelligence Workshop’ as the implementation context. Guided by the principles of emerging engineering education, a student-centered philosophy, and an OBE- and project-driven teaching framework, the study systematically improves the curriculum system. By introducing a blockchain-enabled data governance layer, this platform ensures the tamper-resistance of learning records, enables secure data sharing between universities and enterprises, and facilitates transparent smart-contract-based assessments. This article shows that by carrying out interdisciplinary collaboration and real enterprise case projects on the trusted training platform, students’ digital thinking and algorithm practice abilities have significantly improved, and teachers have also made progress in achievement transformation. Relevant teaching practices indicate that the blockchain-enabled ‘Digital Intelligence Workspace’ effectively promotes the improvement of talent cultivation quality in universities, providing a feasible new approach and practical reference for undergraduate education reform in the era of intelligence.
- 10 August 2026
- Published by Frontiers Media SA in Frontiers in Blockchain
- Vol. 9, 1882200
Open AccessExploring digital skills cultivation through a blockchain-enabled ‘digital-intelligent workshop’
Due to the rapid development of new generation information technology worldwide, universities’ cultivation of digital and intelligent talents has become a necessity of the times. To address the common challenges in traditional practice-oriented teaching, including outdated course design, insufficient authenticity of experimental environments, and cross-institutional trust deficits in student evaluation, this study takes the ‘Digital-Intelligence Workshop’ as the implementation context. Guided by the principles of emerging engineering education, a student-centered philosophy, and an OBE- and project-driven teaching framework, the study systematically improves the curriculum system. By introducing a blockchain-enabled data governance layer, this platform ensures the tamper-resistance of learning records, enables secure data sharing between universities and enterprises, and facilitates transparent smart-contract-based assessments. This article shows that by carrying out interdisciplinary collaboration and real enterprise case projects on the trusted training platform, students’ digital thinking and algorithm practice abilities have significantly improved, and teachers have also made progress in achievement transformation. Relevant teaching practices indicate that the blockchain-enabled ‘Digital Intelligence Workspace’ effectively promotes the improvement of talent cultivation quality in universities, providing a feasible new approach and practical reference for undergraduate education reform in the era of intelligence.
- 10 January 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 336
HFLMND: Toward robust and efficient hierarchical federated learning via malicious node detection
- 10 January 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 336
HFLMND: Toward robust and efficient hierarchical federated learning via malicious node detection
- 2 January 2026
- Published by Springer Nature
- p. 308-321
A Distributed Identity Authentication Scheme for the Metaverse
- 2 January 2026
- Published by Springer Nature
- p. 308-321
A Distributed Identity Authentication Scheme for the Metaverse
- 2 January 2026
- Published by Springer Nature
- p. 288-307
Multi Keyword Dynamic Searchable Encryption Scheme Based on Blockchain
- 2 January 2026
- Published by Springer Nature
- p. 288-307
Multi Keyword Dynamic Searchable Encryption Scheme Based on Blockchain
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 1449-1464
Analyzing Request Volatility in Cloud-Based Machine Learning: Insights From Alibaba’s Machine Learning as a Service Platform
With advancements in machine learning (ML) technology and the deployment of large ML-as-a-Service (MLaaS) clouds, accurately understanding request behaviors in an MLaaS cloud platform is paramount for resource scheduling and optimization. This paper sheds light on the correlation of request arrivals in a representative and dynamic MLaaS workload – Alibaba PAI (an ML platform for artificial intelligence). For requests in the PAI workloads at the job, task, instance, and machine levels, our burstiness diagnosis reveals that the request arrival processes at all levels are significantly bursty. Additionally, our Gaussianity test indicates that the bursty activities in PAI consistently appear to be non-Gaussian. Our findings show that there exists a certain degree of correlation between request arrivals at each level over long-term time scales. Moreover, we reveal the self-similar nature of request activities in the various-level wild MLaaS workloads on Alibaba PAI through visual evidence, the auto-correlation structure of the aggregated process of request sequences, and Hurst parameter estimates. Furthermore, we implement a versatile workload synthetic model to synthesize request series based on the inputs measured from the PAI trace. Experimental results demonstrate that our model outperforms typical self-similar workload models, and can improve accuracy by up to 99% compared to them.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 1449-1464
Analyzing Request Volatility in Cloud-Based Machine Learning: Insights From Alibaba’s Machine Learning as a Service Platform
With advancements in machine learning (ML) technology and the deployment of large ML-as-a-Service (MLaaS) clouds, accurately understanding request behaviors in an MLaaS cloud platform is paramount for resource scheduling and optimization. This paper sheds light on the correlation of request arrivals in a representative and dynamic MLaaS workload – Alibaba PAI (an ML platform for artificial intelligence). For requests in the PAI workloads at the job, task, instance, and machine levels, our burstiness diagnosis reveals that the request arrival processes at all levels are significantly bursty. Additionally, our Gaussianity test indicates that the bursty activities in PAI consistently appear to be non-Gaussian. Our findings show that there exists a certain degree of correlation between request arrivals at each level over long-term time scales. Moreover, we reveal the self-similar nature of request activities in the various-level wild MLaaS workloads on Alibaba PAI through visual evidence, the auto-correlation structure of the aggregated process of request sequences, and Hurst parameter estimates. Furthermore, we implement a versatile workload synthetic model to synthesize request series based on the inputs measured from the PAI trace. Experimental results demonstrate that our model outperforms typical self-similar workload models, and can improve accuracy by up to 99% compared to them.
- 17 October 2025
- Published by Springer Nature in Cluster Computing
- Vol. 28 (16) , 1-14
Task offloading optimization in IRS-assisted multi-tier computing networks
- 17 October 2025
- Published by Springer Nature in Cluster Computing
- Vol. 28 (16) , 1-14
Task offloading optimization in IRS-assisted multi-tier computing networks
- 3 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (24) , 53396-53409
HFLMLD: Enhancing Robustness in Hierarchical Federated Learning With Multiple Layer Defenses
Hierarchical federated learning (HFL) has attracted significant attention for its communication efficiency and cost-effectiveness. However, its distributed nature makes it vulnerable to malicious attacks, particularly in large-scale, zero-trust edge networks where monitoring nodes is challenging. Existing defenses for traditional two-layer federated learning (FL) are insufficient to address the unique cross-layer collusion attacks possible in HFL. To bridge this gap, we introduce HFLMLD, a robust HFL framework with multiple layer defenses. HFLMLD employs a two-stage hierarchical defense strategy to enhance system resilience. At the edge layer, it combines dimensionality reduction-based detection with a dynamic suspicion score mechanism to identify and neutralize malicious clients. At the cloud layer, a normalized weighted aggregation algorithm is employed to counter threats from compromised edge servers. Our extensive experiments on benchmark datasets show that HFLMLD effectively secures HFL systems against complex, multifaceted attacks while maintaining high model performance.
- 3 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (24) , 53396-53409
HFLMLD: Enhancing Robustness in Hierarchical Federated Learning With Multiple Layer Defenses
Hierarchical federated learning (HFL) has attracted significant attention for its communication efficiency and cost-effectiveness. However, its distributed nature makes it vulnerable to malicious attacks, particularly in large-scale, zero-trust edge networks where monitoring nodes is challenging. Existing defenses for traditional two-layer federated learning (FL) are insufficient to address the unique cross-layer collusion attacks possible in HFL. To bridge this gap, we introduce HFLMLD, a robust HFL framework with multiple layer defenses. HFLMLD employs a two-stage hierarchical defense strategy to enhance system resilience. At the edge layer, it combines dimensionality reduction-based detection with a dynamic suspicion score mechanism to identify and neutralize malicious clients. At the cloud layer, a normalized weighted aggregation algorithm is employed to counter threats from compromised edge servers. Our extensive experiments on benchmark datasets show that HFLMLD effectively secures HFL systems against complex, multifaceted attacks while maintaining high model performance.