Course Information

Course Name	Turkish	Hesaplamalı Bilim ve Mühendislikte Güvenilir ve Açıklanabilir Makine Öğrenmesi
	English	Trustworthy and Explainable Mach. Learning in Computational Sci. and Eng.
Course Code	HBM 537E	Credit	Lecture (hour/week)	Recitation (hour/week)	Laboratory (hour/week)
Semester	-	3	3	-	-
Course Language		English
Course Coordinator		Gülşen Taşkın Kaya Gülşen Taşkın Kaya
Course Objectives		Ensure comprehensive understanding of interpretability and explainability in machine learning for clear and transparent model predictions. Train students to critically analyze machine learning model decisions across various applications. Promote the identification and mitigation of biases and ethical issues in machine learning models to ensure fairness and accountability. Prepare students to adapt interpretability techniques to diverse applications domains.
Course Description		General concepts of interpretability and explainability in CSE, explainable methods, model-agnostic techniques, LIME, causality, trustworthy and robust machine learning, OOD generalization, domain adaptation, zero-shot learning, training neural networks against adversarial inputs, feature selection, neural network interpretability, visualization and saliency maps, deep network explanations, perturbation methods, integrated and gradient-based methods, class-activation and layer-wise relevance maps, reverse engineering, black-box models, scalable, robust meta-models, knowledge distillation, global explanation techniques, partial dependence plots, functional decomposition, ANOVA, interpretable models such as linear models, decision trees, GAMs, SHAP, explainable boosting machines, and neural additive models.
Course Outcomes		M.Sc. students who successfully pass this course gain knowledge, skills and competency in the following subjects: 1. Gain a thorough understanding of the core principles and theories behind machine learning interpretability and explainability, along with fundemental concepts of machine learning. 2. Acquire the ability to evaluate, choose, and apply strategies for explaining machine learning models across different contexts. 3. Learn to analyze and enhance the transparency and trustworthiness of machine learning models for ethical and reliable outputs. 4. Develop the skill to adapt machine learning models to new and unseen data environments, increasing their accuracy. 5. Understand and implement methods to assess and reduce biases in models, enhancing their resilience against data variability and adversarial attacks. 6. Build the skill to critically review the outputs and explanations from machine learning models, recognizing their limitations and possible inaccuracies.
Pre-requisite(s)
Required Facilities
Other
Textbook
Other References

Courses

Help

About