arXiv subjects

Abdullah Khan

Publications and source records attributed to Abdullah Khan.

6 recordsLinked to original sources

arXiv2026

PEFT-MedSAM: Efficient Fine-Tuning of Medical Foundation Models for Explainable Skin Lesion Segmentation

Automated segmentation of skin lesions using deep learning models for dermoscopic images can be very helpful in finding melanomas earlier than they would normally be detected. However, most deep learning methods available do not perform well. The aim of this paper is to present a parameter-efficient fine-tuning method called PEFT-MedSAM for adapting the Medical Segment Anything Model (MedSAM) to automatically segment dermoscopic skin lesions. The PEFT-MedSAM method uses only the lightweight mask decoder for training the model while keeping the pre-trained image encoder and prompt encoder frozen. The experiments performed on the ISIC 2018 benchmark dataset shows that PEFT-MedSAM obtains a dice coefficient of .9411 and an intersection over union value of .8918 when compared to both a fully trained U-Net baseline (.8715 dice coefficient) and zero-shot MedSAM inference (.8997 dice coefficient). The external validation of the model using PH2 dataset shows .9467 dice coefficient with +/- .0310 standard deviation. Supportive evidence for these claims include a p-value less than .0001 for Wilcoxon signed rank tests comparing the two datasets and bootstrap-estimated 95% confidence intervals of [.9364,.9447] that represent the estimated range of possible values for the average dice coefficient obtained by repeating the test. To increase clinical trustworthiness, we used Grad-CAM explainability along with a pointing game based evaluation methodology to evaluate the CNN baseline model on the validation set. The results showed that we had an accuracy rate of 98.27% on the validation set of 519 images and confirmed that the model classified regions containing skin lesions.

cs.CV↗

arXiv2024

FaultExplainer: Leveraging Large Language Models for Interpretable Fault Detection and Diagnosis

Machine learning algorithms are increasingly being applied to fault detection and diagnosis (FDD) in chemical processes. However, existing data-driven FDD platforms often lack interpretability for process operators and struggle to identify root causes of previously unseen faults. This paper presents FaultExplainer, an interactive tool designed to improve fault detection, diagnosis, and explanation in the Tennessee Eastman Process (TEP). FaultExplainer integrates real-time sensor data visualization, Principal Component Analysis (PCA)-based fault detection, and identification of top contributing variables within an interactive user interface powered by large language models (LLMs). We evaluate the LLMs' reasoning capabilities in two scenarios: one where historical root causes are provided, and one where they are not to mimic the challenge of previously unseen faults. Experimental results using GPT-4o and o1-preview models demonstrate the system's strengths in generating plausible and actionable explanations, while also highlighting its limitations, including reliance on PCA-selected features and occasional hallucinations.

cs.AI↗

arXiv2021

Untangling Braids with Multi-agent Q-Learning

We use reinforcement learning to tackle the problem of untangling braids. We experiment with braids with 2 and 3 strands. Two competing players learn to tangle and untangle a braid. We interface the braid untangling problem with the OpenAI Gym environment, a widely used way of connecting agents to reinforcement learning problems. The results provide evidence that the more we train the system, the better the untangling player gets at untangling braids. At the same time, our tangling player produces good examples of tangled braids.

cs.LG↗

arXiv2021

Circle graphs (chord interlacement graphs) of Gauss diagrams: Descriptions of realizable Gauss diagrams, algorithms, enumeration

Chord diagrams, under the name of Gauss diagrams, are used in low-dimensional topology as an important tool for studying curves or knots. Those Gauss diagrams that correspond to curves or knots are called realizable. The theme of our paper is the fact that realizability of a Gauss diagram can be expressed via its circle graph. Accordingly, one can define and study realizable circle graphs (with realizability of a circle graph understood as realizability of any one of chord diagrams corresponding to the graph). Several studies contain theorems purporting to prove the fact. We check several of these descriptions experimentally and find counterexamples to the descriptions of realizable Gauss diagrams in some of these publications. We formulate new descriptions of realizable circle graphs and present an elegant algorithm for checking if a circle graph is realizable. We enumerate realizable circle graphs for small sizes and comment on these numbers. Then we concentrate on one type of curves, called meanders, and study the circle graphs of their Gauss diagrams.

math.GT↗

arXiv2020

Kinetic Ballooning Instability of the Near-Earth Magnetotail in Voigt Equilibrium

For a long time, ballooning instabilities have been believed to be a possible triggering mechanism for the onset of substorm and current disruption initiation in the near-Earth magnetotail. Yet the stability of the kinetic ballooning mode (KBM) in a global and realistic magnetotail configuration has not been well examined. In this paper, stability of the KBM is evaluated for the two-dimensional Voigt equilibrium of the near-Earth magnetotail based on an analytical kinetic theory of ballooning instability in the framework of kinetic magnetohydrodynamic (MHD) model, where the kinetic effects such as the finite gyroradius effect, wave-particle resonances, particle drifts motions are included usually through kinetic closures. The growth rate of the KBM strongly depends on the magnetic field line stiffening factor $S$, which is in turn determined by the effects of the trapped electrons, the finite ion gyroradius, and the magnetic drift motion of charged particles. The KBM is unstable in a finite intermediate range of equatorial $β_{eq}$ and only marginally unstable at higher $β_{eq}$ regime for higher $T_e/T_i$ values. The finite ion gyroradius and the trapped electron fraction enhance the stiffening factor that tends to stabilize the KBM in the magnetotail far away from Earth. On the other hand, the current sheet thinning destabilizes KBM in the lower $β_{eq}$ regime and stabilizes KBM in the higher $β_{eq}$ regime.

physics.plasm-ph↗

arXiv2018

Evaluation of Kinetic Ballooning Instability in the Near-Earth Magnetotail

Ballooning instabilities are widely believed to be a possible triggering mechanism for the onset of substorm and current disruption initiation in the near-Earth magnetotail. Yet the stability of the kinetic ballooning mode (KBM) in a global and realistic magnetotail configuration has not been well examined. In this paper, the growth rate of the KBM is calculated from analytical theory for the two-dimensional Voigt equilibrium within the framework of kinetic magnetohydrodynamic (MHD) model. The growth rate of the KBM is found to be strongly dependent on the field line stiffening factor $S$, which depends on the trapped electron dynamics, the finite ion gyroradius, and the magnetic drift motion of charged particles. Furthermore, calculations show that the KBM is unstable in a finite intermediate range of equatorial $β_{eq}$ values and the growth rate dependence on $β_{eq}$ is enhanced for larger $ρ_i$. The KBM stability is further analyzed in a broad range of $k_y$ for different values of ion Larmor radius $ρ_i$ and gradient ratio $η_j \equiv d\ln(T_j)/d\ln(n_j)$, where $T_j$ is the particle temperature and $n_j$ is the particle density. The KBM is found to be unstable for sufficiently high values of $k_y$, where the growth rate first increases to a maximum value and then decreases due to kinetic effects. The $k_y$ at the maximum growth rate decreases exponentially with $ρ_i$. The current sheet thinning is found to enhance the KBM growth rate and the unstable $β_{eq}$ regime in the near-Earth magnetotail.

physics.space-ph↗