Interpretable ML for Scientific Discovery is a 3-day online hands-on workshop designed to help participants build, evaluate, and interpret machine learning models for scientific data analysis.
The workshop focuses on using Scikit-learn, SHAP, and Yellowbrick to understand model performance, identify influential variables, visualize diagnostics, and extract meaningful scientific insights from machine learning predictions.
Aim
The aim of this workshop is to help participants develop practical skills in interpretable machine learning for scientific discovery. Participants will learn how to build ML models, evaluate their reliability, explain model behavior, and connect model outputs with scientific reasoning.
What Participants Will Learn
Understand the role of machine learning in scientific discovery.
Frame scientific datasets as ML problems.
Build baseline models using Scikit-learn.
Evaluate model performance and reliability.
Understand feature importance and permutation importance.
Use Yellowbrick for model diagnostics and visualization.
Apply SHAP for global and local model explanations.
Extract scientific insights from interpretable ML results.
Structure
📅 Day 1: Machine Learning Foundations for Scientific Data Analysis
Role of machine learning in scientific discoveryUnderstanding scientific datasets, variables, features, and labelsFraming scientific problems as ML tasksData preprocessing and train-test split workflowBuilding baseline models using Scikit-learnModel evaluation and reliability interpretationUnderstanding how ML results support scientific decision-making🛠️ Hands-on:Hands-on 1: Load and prepare a scientific-style dataset using Scikit-learnHands-on 2: Build and evaluate a baseline ML model using Scikit-learn🧰 Tools Covered: Scikit-learn, Python, Jupyter Notebook / Google Colab
📅 Day 2: Feature Importance and Visual Model Interpretation
Introduction to interpretable machine learningUnderstanding feature importance in scientific modelsBuilt-in feature importance vs permutation importanceUsing permutation importance to identify influential variablesYellowbrick visualizations for model diagnosticsInterpreting influential variables in scientific datasetsUsing visualization to support model reliability and interpretation🛠️ Hands-on:Hands-on 1: Perform feature importance analysis using Scikit-learnHands-on 2: Apply permutation importance and create Yellowbrick visualizations🧰 Tools Covered: Scikit-learn, Yellowbrick, Python, Jupyter Notebook / Google Colab
📅 Day 3: SHAP Explainability and Scientific Insight Extraction
Introduction to SHAP for explainable AILocal vs global model interpretationSHAP summary plots for global explanationsSHAP explanations for individual predictionsConnecting model explanations with scientific reasoningGuided final exercise for scientific insight extractionPreparing a short scientific interpretation from model explanations🛠️ Hands-on:Hands-on 1: Generate SHAP explanations for global model interpretationHands-on 2: Interpret individual predictions and prepare a short scientific insight summary🧰 Tools Covered: SHAP, Scikit-learn, Yellowbrick, Python, Jupyter Notebook / Google Colab
Important Dates
Registration Ends
4.30 pm
Workshop Dates
2026-06-02
5.30 PM
5.30 PM
What You Will Gain
Live & recorded sessions
e-Certificate upon completion
Post-workshop query support
Hands-on learning experience
Outcomes
Prepare scientific-style datasets for ML workflows.
Build baseline machine learning models using Scikit-learn.
Evaluate ML models for reliability and performance.
Interpret feature importance in scientific datasets.
Apply permutation importance for robust model interpretation.
Use Yellowbrick visualizations for model diagnostics.
Generate SHAP explanations for global and local model interpretation.
Connect ML explanations with scientific reasoning.
Prepare a short scientific insight summary based on model results.
Who Should Attend
PhD scholars and research scholars working with scientific datasets
Postgraduate students interested in machine learning for research
Academicians and faculty members exploring interpretable AI
Researchers from science, engineering, biotechnology, healthcare, environment, or materials domains
Industry R&D professionals using data-driven models
Data science beginners interested in explainable AI
Participants who want to understand how ML models make predictions
Basic knowledge of Python or machine learning is helpful, but advanced expertise is not mandatory.
