Interests
Publications
Bidirectional xLSTM with a Kolmogorov–Arnold Emission Head for Bangla Named Entity Recognition
Under Review, ICCIT 2026S. Yeasar, et al.
A BanglaBERT → bidirectional xLSTM → CRF architecture with a Kolmogorov–Arnold Network (KAN) emission head — the first application of xLSTM to Bangla NER. Attains 85.30 ± 0.06 entity-level macro F1 on BanNERD, outperforming BiLSTM–CRF and CRF baselines.
R. R. Rudra, S. Yeasar, T. A. Noor, et al.
Applied Random Forest and SHAP analysis to evaluate the drought tolerance of five major Bangladeshi crops using 16 climatic and geographic variables.
L. C. Roy, S. K. Das, M. S. Yeasar, et al.
Developed an ARIMA(2,1,2)-based rainfall forecasting model using 40+ years of meteorological data from 30 stations across Bangladesh. Presented at the 7th International Conference on Advances in Civil Engineering, CUET, Bangladesh.