Scientific Machine Learning
Scientific Machine Learning (SciML) integrates machine learning techniques with scientific computing, physics, and mathematical modelling. My research applies machine learning approaches to ecological systems, wetland vulnerability assessment, dynamical systems, and biological modelling.
The objective is to combine data-driven learning with mechanistic understanding to produce accurate, interpretable, and computationally efficient predictive models. Research in this area also explores uncertainty quantification, hybrid modelling, and scientific data assimilation.
Methodological Framework
- Physics-guided machine learning
- Bayesian inference and uncertainty quantification
- Hybrid mechanistic-data-driven modelling
- Deep learning for dynamical systems
- Environmental vulnerability prediction
- Ecological data analysis and forecasting
Applications
- Wetland habitat vulnerability analysis
- Ecological sustainability assessment
- Pattern recognition in dynamical systems
- Neurodegenerative disease modelling
- Environmental risk prediction
Research Highlights
Machine learning models have been used to predict wetland habitat vulnerability in the mature Ganges delta, combining environmental indicators with AI-based predictive frameworks. The work demonstrates the potential of hybrid machine learning approaches for environmental sustainability and ecological forecasting.
Selected References
1. Pal, S., & Debanshi, S. (2021). Machine learning models for wetland habitat vulnerability in mature Ganges delta. Environmental Science and Pollution Research.
2. Pal, S., & Melnik, R. (2025). Adaptive modelling using Bayesian PINNs for Alzheimer's disease.
3. Pal, S., Banerjee, M., & Melnik, R. (2024). Nonlocal Models in Biology and Life Sciences.