Explainable Machine Learning
General data
Course ID: | 1000-1M18WUM |
Erasmus code / ISCED: |
11.1
|
Course title: | Explainable Machine Learning |
Name in Polish: | Wyjaśnialne uczenie maszynowe |
Organizational unit: | Faculty of Mathematics, Informatics, and Mechanics |
Course groups: |
(in Polish) Przedmioty obieralne na studiach drugiego stopnia na kierunku bioinformatyka Elective courses for 2nd stage studies in Mathematics Elective courses for Computer Science |
ECTS credit allocation (and other scores): |
(not available)
|
Language: | English |
Type of course: | elective monographs |
Short description: |
To learn the objectives, methods and techniques of explaining complex machine learning models, the black box models. Predictive models are increasingly complex, tree committees, deep neural networks are models with thousands of parameters. For models with such dimensions, it is easy to lose control over what the model has learned. In this subject we will discuss tools to analyze the structure of the model treated as a black box, and to analyze the predictions from this model. This will increase the confidence in the model, improve the efficiency of the model, and the ability to draw useful knowledge from the model. |
Full description: |
Lecture: Understanding of the model: - Measures of identifying important variables (based on permutations, based on loss functions), - model quality testing measures (for regression and classification models), - measurements of the boundary model response (partial model response, conditional model response, individual model responses). Understanding of predictions: - local approximations and the LIME approach, - attribution of the importance of features based on breakDown and Shapley values method. Laboratory: Performing predictive analysis for a specific phenomenon. Application of methods of explaining for a given phenomenon. Project: Implementation of a new library or validation of the chosen algorithm of understanding black box models. |
Bibliography: |
1. Examples and documentation for Descriptive mAchine Learning EXplana-tions. Biecek 2018. https://pbiecek.github.io/DALEX_docs 2. Ribeiro, Marco Tulio, Sameer Singh, and Carlos Guestrin. 2016. “‘Why Should I Trust You?’: Explaining the Predictions of Any Classifier.” In, 1135–44. ACM Press. https://doi.org/10.1145/2939672.2939778. 3. Fisher, Aaron, Cynthia Rudin, and Francesca Dominici. 2018. “Model Class Reliance: Variable Importance Measures for Any Machine Learning Model Class, from the ’Rashomon’ Perspective.” Journal of Computational and Graphical Statistics. http://arxiv.org/abs/1801.01489. |
Learning outcomes: |
KNOWLEDGE W01 Knows basic methods of data pre-treatment, including data size reduction and feature extraction. W02 Knows basic methods of XAI and their use in business data analysis SKILLS U01 Knows the basic methods of studying the structure of ML models and their use in business data analysis. U02 is able to build a classifier and assess the significance of individual variables for the final result. SOCIAL COMPETENCES K01 Able to work in a project group taking on different roles in it |
Assessment methods and assessment criteria: |
The evaluation will consist of three components activity during classes (20%), housework (20%) project (60%) You need at least 50% of the points to pass. |
Copyright by University of Warsaw.