Ali Eslamian

image

I am a PhD student in Computer Science at the University of Kentucky , working at the intersection of machine learning, tabular data modeling, bioinformatics, and neuroimaging.

My research focuses on developing scalable and interpretable deep learning models for structured and multimodal data, with applications in Alzheimer’s disease and dementia research.

Research Interests

  • Tabular Deep Learning
  • Interpretable Machine Learning
  • Sparse Attention Models
  • Neuroimaging and Biomarker Discovery
  • Alzheimer's Disease and Dementia
  • Bioinformatics and Multimodal Learning
  • Pattern Recognition and Computer Vision

Education

  • Ph.D. in Computer Science
    University of Kentucky, Lexington, KY, USA
    January 2024 - Present
  • M.Sc. in Electrical Engineering (Communication Systems)
    Isfahan University of Technology, Iran
    October 2020 - June 2023
  • B.Sc. in Electrical Engineering
    University of Isfahan, Iran
    October 2016 - September 2020

Publications

TabNSA: Native Sparse Attention for Efficient Tabular Data Learning
2026

Journal of Neurocomputing, 2026 • Ali Eslamian, Qiang Cheng

Sparse attention for efficient feature interaction modeling. Targets scalability on high-dimensional tabular datasets. Designed to balance performance and efficiency.
📄 Paper 💻 Code
Interpretable AI-Driven Biomarker Kits for Early Detection of Dementia using Multimodal SCAN Data
2025

AAIC 2025 • Ali Eslamian, Qiang Cheng, Colleen Pappas, Christopher E. Bauer, Brian T. Gold

Multimodal learning for dementia biomarker discovery. Focus on interpretability and clinically meaningful signals. Designed to support early detection using SCAN data.
📄 Paper
TabKAN: Advancing Tabular Data Analysis using Kolmogorov-Arnold Network Learning
2025

Journal of Machine Learning for Computational Science and Engineering, 2025 • Ali Eslamian, Afzal Aghaei, Alireza, Qiang Cheng

Interpretable tabular learning via Kolmogorov–Arnold Networks. Improves modeling of feature transformations and interactions. Targets robustness across diverse tabular datasets.
📄 Paper 💻 Code
TabMixer: Advancing Tabular Data Analysis with an Enhanced MLP-Mixer Approach
2025

Journal of Pattern Analysis and Applications, 2025 • Ali Eslamian, Qiang Cheng

Enhanced MLP-Mixer for structured tabular representation learning. Designed for scalable training and strong baselines. Supports supervised and transfer learning settings.
📄 Paper 💻 Code
Evaluating Deep Learning Models for Breast Cancer Classification: A Comparative Study
2025

SPIE Medical Imaging 2025 • Sania Eskandari, Ali Eslamian, Nusrat Munia, Amjad Alqarni, Qiang Cheng

Comparative evaluation of deep models for breast cancer classification. Emphasis on practical performance and rigorous benchmarking. Medical imaging-focused analysis and reporting.
📄 Paper 💻 Code
Improving SCGAN’s Similarity Constraint and Learning a Better Disentangled Representation
2024

International Conference on Electrical Engineering 2024 • Iman Yazdanpanah, Ali Eslamian

Improves similarity constraint design in SCGAN. Targets better disentanglement and representation quality. Validated with experimental results in the paper.
📄 Paper
Det-SLAM: A Semantic Visual SLAM for Highly Dynamic Scenes Using Detectron2
2022

IEEE ICSPIS 2022 • Ali Eslamian, Mohammad Reza Ahmadzadeh

Semantic SLAM for dynamic scenes using Detectron2. Reduces impact of moving objects on localization accuracy. Designed for robust robot perception in real environments.
📄 Paper 💻 Code
A Novel Approach to EKF-SLAM in Dynamic Environments: Euclidean Distance-Based Landmark Detection
Preprint

SSRN Preprint • Ali Eslamian, Masoud Dorvash, Mohammad Reza Ahmadzadeh

Euclidean distance-based methods to mitigate moving landmarks. Applied to EKF-SLAM and UKF-SLAM in dynamic settings. Validated via simulations and real-world scenarios.
📄 Paper 💻 Code

Current Research

  • Scalable and Interpretable Deep Learning for Tabular Data
    Development of TabMixer, TabNSA, and TabKAN models for supervised, transfer, and feature-incremental learning on structured data.
  • AI-Driven Biomarker Kits for Alzheimer's Disease
    Multimodal learning using MRI, plasma biomarkers, genetics, and clinical data from NACC, ADNI, and UK-ADRC cohorts.