arXiv subjects

Lilatul Ferdouse

Publications and source records attributed to Lilatul Ferdouse.

3 recordsLinked to original sources

arXiv2026

Uncertainty-Aware and Explainable Ensemble Deep Learning Framework for Multi-Class Skin Lesion Classification

Skin cancer diagnosis from dermoscopic images remains challenging due to high intra-class variability, inter-class similarity, class imbalance, and the limited interpretability of deep learning models. This paper proposes an uncertainty-aware and explainable deep learning framework for multi-class skin lesion classification. The framework combines a vision transformer model (MaxViT-Tiny) with CNN-based models (ConvNeXt-Tiny and EfficientNetV2-B0) through deep ensemble learning. Monte Carlo (MC) Dropout estimates predictive uncertainty and identifies unreliable predictions, while Grad-CAM++, an explainable AI (XAI) technique, provides visual explanations by highlighting lesion regions that influence model decisions. Evaluated on the HAM10000 dataset, the framework achieves 96% accuracy and 99% ROC-AUC under uncertainty-aware filtering (entropy < 1.0, confidence >= 0.7), with macro-average precision, recall, and F1-score of 94%, 95%, and 95%, respectively, and 96% weighted-average scores across all three metrics. The results demonstrate accurate, interpretable, and uncertainty-aware skin lesion classification for trustworthy computer-aided diagnosis.

eess.IV↗

arXiv2011

Implementation of the Trigonometric LMS Algorithm using Original Cordic Rotation

The LMS algorithm is one of the most successful adaptive filtering algorithms. It uses the instantaneous value of the square of the error signal as an estimate of the mean-square error (MSE). The LMS algorithm changes (adapts) the filter tap weights so that the error signal is minimized in the mean square sense. In Trigonometric LMS (TLMS) and Hyperbolic LMS (HLMS), two new versions of LMS algorithms, same formulations are performed as in the LMS algorithm with the exception that filter tap weights are now expressed using trigonometric and hyperbolic formulations, in cases for TLMS and HLMS respectively. Hence appears the CORDIC algorithm as it can efficiently perform trigonometric, hyperbolic, linear and logarithmic functions. While hardware-efficient algorithms often exist, the dominance of the software systems has kept those algorithms out of the spotlight. Among these hardware- efficient algorithms, CORDIC is an iterative solution for trigonometric and other transcendental functions. Former researches worked on CORDIC algorithm to observe the convergence behavior of Trigonometric LMS (TLMS) algorithm and obtained a satisfactory result in the context of convergence performance of TLMS algorithm. But revious researches directly used the CORDIC block output in their simulation ignoring the internal step-by-step rotations of the CORDIC processor. This gives rise to a need for verification of the convergence performance of the TLMS algorithm to investigate if it actually performs satisfactorily if implemented with step-by-step CORDIC rotation. This research work has done this job. It focuses on the internal operations of the CORDIC hardware, implements the Trigonometric LMS (TLMS) and Hyperbolic LMS (HLMS) algorithms using actual CORDIC rotations. The obtained simulation results are highly satisfactory and also it shows that convergence behavior of HLMS is much better than TLMS.

cs.OH↗

arXiv2011

Simulation and Performance Analysis of Adaptive Filtering Algorithms in Noise Cancellation

Noise problems in signals have gained huge attention due to the need of noise-free output signal in numerous communication systems. The principal of adaptive noise cancellation is to acquire an estimation of the unwanted interfering signal and subtract it from the corrupted signal. Noise cancellation operation is controlled adaptively with the target of achieving improved signal to noise ratio. This paper concentrates upon the analysis of adaptive noise canceller using Recursive Least Square (RLS), Fast Transversal Recursive Least Square (FTRLS) and Gradient Adaptive Lattice (GAL) algorithms. The performance analysis of the algorithms is done based on convergence behavior, convergence time, correlation coefficients and signal to noise ratio. After comparing all the simulated results we observed that GAL performs the best in noise cancellation in terms of Correlation Coefficient, SNR and Convergence Time. RLS, FTRLS and GAL were never evaluated and compared before on their performance in noise cancellation in terms of the criteria we considered here.

cs.OH↗