Methodology gaps in Computer Science

Representative open questions

Showing 30 of 479 — one per source paper, highest-quality first.

• Federated learning for privacy-preserving skin cancer classification using deep neural networks (2026) · doi

  The federated learning approach using the ring topology outperforms centralized training on the ISIC dataset, but the effect of different topologies (e.g., FedAvg, FedProx) is not fully explored.

• Causal K-means clustering (2026) · doi

  The causal K-means clustering method requires separate independent samples for constructing the nuisance parameter estimator μ̂ to satisfy Assumption A2; the paper does not explore whether the sample-splitting requirement can be relaxed or how performance degrades when a single sample is used for both clustering and nuisance estimation.

• AN ITERATIVE GLMM–XGBOOST ALGORITHM WITH GROUP-AWARE CONDITIONAL PERMUTATION IMPORTANCE FOR EXPLAINING MULTILEVEL ITEM RESPONSE DATA (2026) · doi

  The conditional GLMM–XGBoost advantage over conditional GLMM is described as 'comparatively modest' with empirical Bayes predictions incorporated in both, but the paper does not isolate or quantify the specific contribution of the XGBoost non-parametric component versus random-effects conditioning to this improvement.

• Cross-Currency Basis Swaps Referencing Backward-Looking Rates (2026) · doi

  The pricing framework explicitly handles multi-period CCBS with tenor structure T_n, but numerical implementation details for computing the nested expectations in terms I₂ₜ, I₃ₜ, and I₄ₜ (which involve products of multiple Gamma functions with correlated parameters) are not provided. Computational complexity and convergence criteria for Monte Carlo or tree-based approximations remain unaddressed.

• Securing IoT Devices with PUFs: Mitigating Aging and Tampering through Cryptography and Machine Learning (2026) · doi

  The linear complexity test requires M to be chosen between 500 and 5000 for the limiting distribution to provide reasonable approximation, but the paper does not specify how to select the optimal M value within this range for PUF-based cryptographic key generation in IoT devices with varying hardware constraints.

• Blockchain-integrated machine learning framework for transparent smart contract vulnerability detection (2026) · doi

  The confusion matrices show that XGBoost commits errors between neighboring vulnerability classes (denial_of_service and arithmetic) more frequently than CatBoost and Random Forest, but the feature space geometry and decision boundary characteristics causing these misclassifications in smart contract vulnerability detection have not been examined. Analyzing the geometric properties of feature representations for confused vulnerability pairs would clarify this phenomenon.

• Can deep learning-based segmentation and classification improve the detection of renal cortical abnormalities? (2026) · doi

  The paper does not investigate how CLAHE preprocessing parameters (clip limit, tile grid size) were optimized or whether these parameters are transferable across different DMSA imaging equipment and acquisition protocols. Future work should systematically evaluate CLAHE preprocessing robustness for DenseNet205 classification across DMSA systems with varying imaging parameters and reconstruction algorithms.

• Inferring High-Dimensional Dynamic Networks Changing with Multiple Covariates (2026) · doi

  The distinction between sparse edges present in all nine subgraphs (core network unaffected by covariates) versus covariate-influenced edges is determined by the tuning parameters γ1 and γ2 (set at 5×10⁻⁵ and 5×10⁻⁶), but systematic guidelines for selecting these penalty parameters in CVN estimation across different biological datasets and covariate structures are not provided.

• Umjetna inteligencija: od kritičkih promišljanja do njezine primjene u kaznenom pravosuđu (2026) · doi

  The Italian Delia algorithm processes over 1.5 million variables combining geographic crime data with witness interviews and CCTV footage, but no comparative analysis exists documenting which specific variables contribute most to prediction accuracy or how the system performs across different geographic regions and crime categories.

• Prediction of sedimentation concentration profiles in inclined suspension systems: A data-driven neural network framework (2026) · doi

  The gamma-ray attenuation measurements demonstrate localized relative errors during the passage of the settling front where concentration–time curves exhibit steep gradients; methods to reduce these transient region discrepancies in spatiotemporal concentration profiles using advanced signal processing or improved ANN architectures (e.g., physics-informed neural networks) have not been evaluated.

• Large language model based machine translation for universal multilingual understanding and translation quality enhancement (2026) · doi

  The paper identifies that smaller specialized models like ALMA-13B-LoRA are more effective for low-resource and exotic languages such as Icelandic, but does not provide systematic investigation of which architectural features or fine-tuning strategies enable this superiority for low-resource language translation in multilingual LLMs.

• Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry (2026) · doi

  The supervised stacked autoencoder achieved 91.1% overall accuracy, but performance degradation on chemically complex classes (feldspar cSA, feldspar cSOA, cellulose) with high reconstruction errors has not been systematically addressed. Architectural modifications to the autoencoder that explicitly handle compositionally heterogeneous spectra with elevated SSE values require investigation.

• Quantum Information Framework for Neural Network Generalization: A Comprehensive Experimental Analysis (2026) · doi

  Weight entropy is computed using histogram binning with a fixed bin count of 50; the sensitivity of classical weight entropy calculations to bin selection and alternative entropy estimators (e.g., kernel density estimation) in the context of quantum information framework validation has not been addressed.

• AI in Cybersecurity: A Systematic Review and Conceptual Audit Model (2026) · doi

  The model proposes that AI-enhanced auditing should incorporate human judgment into the multi-layered audit process, but does not specify the decision points where human expertise is required, the criteria for human override of AI recommendations, or how to validate the quality of human-AI collaborative decisions in cybersecurity auditing.

• Research on a strongly generalizable fault diagnosis method based on adversarial transfer learning (2026) · doi

  While HDAL combines DANN's adversarial strategy with DSAN's LMMD subdomain alignment for fine-grained fault feature alignment, the paper does not investigate whether additional domain alignment objectives (e.g., class-aware alignment, partial domain adaptation) could further improve performance on the difficult-to-distinguish Conditions 12 and 13 in cross-reactor adversarial transfer learning.

• Machine-learning-based reconstruction of Ming-dynasty defensive corridors in Yuxian (2026) · doi

  The comprehensive cost surface construction weights multiple single-factor cost functions through weighted summation, but the paper does not specify the weighting scheme methodology, sensitivity analysis of weight values, or whether different weighting strategies (inverse distance weighting, analytical hierarchy process) produce significantly different corridor network predictions.

• Comparative analysis of deep learning algorithms for rolling element bearing fault classification under variable loads and speeds (2026) · doi

  The paper observed an abrupt loss climb between epochs 8-12 during CNN training (loss reaching 20-25 with accuracy dropping to 0.2-0.3), followed by stabilization, with this behavior varying significantly across different fold splits. Future investigation should characterize the causes of this non-monotonic training behavior in bearing fault classification networks and develop regularization or optimization strategies to prevent or mitigate these training instabilities.

• LLM-Powered Silent Bug Fuzzing in Deep Learning Libraries via Versatile and Controlled Bug Transfer (2026) · doi

  The repeated validation analysis (Fig. 11b) examines LLM stability across multiple checks per prompt for silent bug classification, but the paper does not specify how many repeated checks are empirically necessary to reach stable predictions or how performance degrades with incomplete API documentation across different deep learning libraries. Determining optimal repetition thresholds for reliable silent bug confirmation is needed.

• Quantum-SpinalNet: a hybrid deep learning approach for mammographic breast cancer detection (2026) · doi

  While the paper claims quantum-inspired entanglement and superposition principles in the DQNN improve feature interaction modeling, there is no comparative analysis of the computational complexity, inference time, or quantum advantage of the DQNN classifier versus classical deep neural network classifiers on the same mammographic segmentation task.

• Ocean: Object-aware Anchor-free Tracking with Matching-relation Learning (2026) · doi

  The comparison with unsupervised learning-based visual trackers on TrackingNet shows inferior performance, but no investigation is provided into whether the matching-relation learning approach can be extended with unsupervised or self-supervised components to reduce annotation requirements while maintaining anchor-free tracking accuracy.

• From unstructured text to structured reasoning: a hybrid knowledge graph for Indonesian sentencing analysis (2026) · doi

  Evidence entity extraction achieves 87.1-89.4% F1-score with acknowledged boundary ambiguity between items (e.g., surveillance video vs. video evidence); the paper does not propose or test methods to resolve these boundaries through hierarchical evidence ontologies or context-dependent entity linking in the knowledge graph.

• Neural Network Tools in the Arsenal of a University Teacher (2026) · doi

  The paper identifies that AI-based tools perform worse than traditional methods at precision in identifying relevant articles but better at finding unique publications, yet provides no empirical framework for measuring or optimizing the trade-off between relevance accuracy and uniqueness discovery in hybrid AI-traditional search methodologies for academic literature.

• Predicting Employee Attrition: A Machine Learning Approach in Human Resource Analytics (2026) · doi

  The paper identifies that managerial relationships (Years with Current Manager) contribute significantly to attrition predictions, but does not investigate whether manager-specific characteristics (leadership style, retention rate of their team, tenure in role) would improve predictive performance beyond tenure-based features in the employee attrition model.

• Enhancing Breast Cancer Diagnosis through Machine Learning: A Robust Approach for Early Detection (2026) · doi

  Advanced dimensionality reduction techniques such as Principal Component Analysis (PCA) and t-SNE to reduce feature complexity and mitigate overfitting in the breast cancer machine learning models have been proposed but not yet applied or evaluated for their impact on model performance and efficiency.

• DAFRL: a dynamic adaptive mean field game-based multi-agent cooperative decision-making method (2026) · doi

  The design of action space and state observation for three-dimensional UAV swarm scenarios requires systematic optimization. Current validation uses 2D state representations and discrete action spaces defined for planar interception; extension to 6-DOF agent dynamics with continuous action spaces needs investigation regarding observability and computational scalability.

• Scour depth prediction using machine learning and explainable AI: assessment of bridge vulnerability (2026) · doi

  Support Vector Regression (SVR) consistently underperformed with RRMSE values up to 0.5685 and R² values of 0.6528–0.841, showing highest sensitivity to input variability; the paper does not investigate whether alternative SVR kernel functions, hyperparameter tuning strategies beyond standard optimization, or kernel-based feature engineering could improve SVR's capability for capturing nonlinear scour processes.

• A Motion-Based Compression and Tracking System for Video Camera Trap-Based Insect Behaviour Studies (2026) · doi

  The motion-based compression algorithm currently uses foreground changes as motion triggers without discriminating the cause of motion, resulting in false (non-insect) detections caused by wind and illumination changes in dynamic monitoring environments. The paper identifies that assessing foreground change-based motion triggers using deep learning prior to storing video frames could minimize these false detections, but this approach requires investigation of the trade-offs between improved compression efficiency and increased computational overhead on resource-constrained edge devices.

• A general framework for Gaussian Splatting-based human-centric volumetric videos (2026) · doi

  The training process for 3D Gaussian Splatting still relies exclusively on stochastic gradient descent (SGD) optimization, which exhibits linear convergence and hyperparameter sensitivity bottlenecks, particularly evident with long sequences or high-resolution inputs where training times remain considerable and impractical for real-time applications.

• Minimum Stable Cut and Treewidth (2026) · doi

  The Propagator gadget (Step 3) enforces consistency across consecutive Selector gadgets but the paper does not explore whether alternative propagation mechanisms or fewer Propagator instances could reduce the overall construction size or improve the hardness lower bound dependence on k and m.

• Sensor Data Fusion in Healthcare Monitoring System with Appropriate Rule-based Model for Error Reduction (2026) · doi

  While the paper demonstrates that sensor fusion reduces false alarm rate from 1.8% (temperature sensor alone) to 1.2% (fused system), it does not specify the rule-based decision mechanisms or thresholds used in the RSF-HMM-SDL model to achieve this error reduction, limiting reproducibility of the fusion algorithm.