Skin cancer is one of the most common malignancies in the world and early and accurate dermoscopic diagnosis is crucial for better survival outcomes of patients. There are various limitations in current single model convolutional and transformer models, such as limited ability to capture local texture, multi-scale morphology, and global contextual information, as well as high intra-class visual similarity and extreme class imbalance. To overcome these problems, this paper proposes an Attention-Guided Ensemble Deep Learning (AGEDL) framework which integrates the EfficientNet-B3, InceptionV3 and Swin Transformer to simultaneously learn complementary dermoscopic representations from the seven-class ISIC 2018 dataset. The Squeeze-and-Excitation (SE) attention blocks dynamically modulate feature responses in each channel, which reduces background noise and enhances discriminative features of lesions. The Lion optimizer offers stable and efficient training in both base training and fine-tuning stages. An XGBoost stacking metalearner is used to combine all backbone networks, and hyperparameters of the XGBoost are optimized by the Dhole Optimizer. The proposed AGEDL achieved 98.02% accuracy, 97.09% precision, 97.05% recall and 94.07% F1 score with 95% confidence interval of 97.41%–98.59%, surpassing the state-of-the-art CNN, transformer-based and hybrid ensemble baselines.
Dharavath Nagesh, Erukonda Jairam· 2026 4th International Confe...· 0 citations
This research proposes a new framework of analysis in predicting high-impact deviations in an environmental system in changing observational circumstances. The new approach, which is called Distribution-adaptive Uncertainty Synthesis (DAUS) is made to work without explicit physical assumptions or time-order-dependence, addressing instead latent structure regularities in the heterogeneous observations. DAUS combines regime invariant encoding and uncertainty resilient optimization to align with sub-surface distributional instability which is a precursor to extreme behavior of a system. The structure utilizes the dual-channel latent representations in maintaining variability and structural consistency and adaptive uncertainty synthesis mechanism dynamically increases signals linked to high deviation potential. A self-recalibration thresholding approach also allows the end-on continuous recalibration of non-stationary input distributions. In comparison to traditional predictive structures, DAUS is based on anticipatory sensitivity instead of having point estimation accuracy, enabling it to be practical in case of sudden regime changes and partial information. Experimental assessment on various benchmark data proves that the suggested strategy always yields superior results compared to current strategies in detecting high-deviation cases, especially those in the cases of distributional volatility and the presence of noise. These findings suggest that DAUS is a good and generalizable route to further study of the environmental system and has significant potentials of being integrated into decision-support pipelines where the uncertainty awareness and adaptive responsiveness is essential. The suggested technique attains an overall accuracy of roughly 91.6%, indicating its robust and equitable performance across detection reliability metrics.
Dharavath Nagesh, P. Deepthi, A. Sahu et al.· 2026 5th OPJU International...· 0 citations
