Covers: application of Monotonic Constraints
Estimated time needed to finish: 60 minutes
Questions this item addresses:
- What is an example of applying monotonic constraints to migitge bias?
How to use this item?
This paper highlights the risk that non-linear machine learning models can learn rules that violate ethical principles and social norms. For example, years of experience can be a proxy for age and penalizing job candidates for having more experience can constitute a form of age discrimination. The paper illustrates how monotonicity shape constraints to can be used to remediate model violations that can cause social harm.
Author(s) / creator(s) / reference(s)
Serena Wang; Maya Gupta
0 comment
Monotonic Constraints, Interaction Constraints and Cost-Sensitive Learning
Contributors
- Objectives
- Techniques for placing constraints on machine learning models to ensure they don't end up with illogical or biased results
- Potential Use Cases
- Bias-mitigation and for injecting domain knowledge into the model
- Who is This For ?
- INTERMEDIATEMachine Learning Developers focused on bias mitigation and model validation
Click on each of the following annotated items to see details.
ARTICLE 1. Bias in Machine learning models
- What is bias and sources of bias in ML models?
10 minutes
PAPER 2. Deontological Ethics By Monotonicity Shape Constraints
- What is an example of applying monotonic constraints to migitge bias?
60 minutes
ARTICLE 3. Application of Monotonic Constraints in Machine Learning Models
- How to implement monotonic constraints on a Machine Learning Model?
15 minutes
REPO 4. Feature Interaction Constraints in XGBoost
- What is a code example of applying feature interaction constraints on an XGBoost model?
60 minutes
PAPER 5. Feature Interactions in XGBoost
- What is feature interaction in non-linear models?
34 minutes
PAPER 6. Cost-Sensitive Learning and the Class Imbalance Problem
- What is the motivation for cost sensitive learning?
60 minutes
Concepts Covered
0 comment