Projects

This project develops a Hybrid-AI Quantile Regression framework, merging nonparametric statistical methods with physics-informed neural networks.
It aims to enhance the understanding and prediction of climate phenomena by analyzing spatio-temporal climate data.
The framework’s versatility is showcased in a case study using rainfall data from India and Israel, highlighting its potential in environmental research, urban planning, and climate change mitigation.

This project aims to develop innovative methods for model selection and lack-of-fit testing in the context of Ordinary Differential Equations (ODEs) modeling of dynamical systems.
Focusing on nonlinear systems, both fully and partially observed, it introduces testing methods for model selection and nonparametric testing of lack-of-fit.
This research will have significant implications in various scientific fields, such as engineering, medicine, psychotherapy, epidemiology and biology, offering new statistical tools for understanding complex dynamical processes.

This project focuses on the development of postharvest protocols using physics-informed machine learning for assessing the quality of cucumbers, zucchini, champignon, and portobello mushrooms under extreme climatic conditions.
By integrating advanced AI with physics-based insights, the project aims to accurately predict produce fitness through image analysis.
In colaboration with MIGAL – Galilee Research Institute.

This project applies dynamical system modeling and statistical inference to enhance understanding and efficacy in psychotherapy treatment.
It employs mechanistic models such as Ordinary Differential Equations (ODEs) and physics-informed machine learning to delve into the complexities of psychotherapeutic processes.
Additionally, mixed effects nonlinear modeling is utilized to capture the nuanced variations in therapy outcomes, offering a comprehensive and sophisticated approach to analyzing and improving psychotherapy practices.

Collaborator: Dr. Amnon Gil, Head of Nephrology & Hypertension Institute, Emek Medical Center
Clinical Need: Despite major technological progress, dialysis patients continue to experience high morbidity and mortality. Personalizing fluid removal is critical to reducing complications such as fluid overload, cardiovascular strain, and infections.
Objective: Develop machine learning and statistical models to personalize ultrafiltration rates (UFR) in hemodialysis patients, with the goal of improving safety and clinical outcomes.
Approach: Combine advanced statistical methods (e.g., regression, mixed-effects modeling) with machine learning techniques (e.g., random forests, neural networks).
Outcome: Creation of predictive models that recommend optimal UFR, tailored to individual patients, balancing effective fluid removal with minimized risks of hypotension and discomfort.