Explainable AI- A New Paradigm for Transparency in AI

Presented at "One Week Workshop on the Internet of Things (IoT)" under the ATAL Program Sponsored by AICTE and organized by "The Department of Computer Science & Technology (Central University of Jharkhand, Ranchi)"

Transcript

1. Explainable AI- A New Paradigm for Transparency in AI Shadab Hussain Data Scientist https://shadabhussain.com/
2. The Problem with Machine Learning
3. The Problem with Machine Learning
4. Trouble with Bias in ML Systems
5. Impact of Bias in ML Systems
6. Impact of Bias in ML Systems
7. Impact of Bias in ML Systems
8. Impact of Bias in ML Systems
9. Impact of Bias in ML Systems
10. Impact of Bias in ML Systems
11. Model Interpretation Strategies Interpretability also popularly known as human interpretable interpretations (HII) of a machine learning model is the extent to which a human (including non-experts in ml) can understand the choices taken by models in their decision making processes.
12. Understanding Model Interpretation • What drives model predictions? • Why did the model take a certain decisions? • How can we trust model predictions?
13. Criteria for Model Interpretation Methods • Intrinsic or Post hoc? • Model Specific or Model Agnostic? • Local or Global?
14. Scope of Model Interpretation
15. Traditional Techniques for Model Interpretation • Exploratory Analysis and Visualization • Model Performance Evaluation Metrics • Use Interpretable Models
16. The Accuracy vs Interpretability Trade-Off
17. Techniques for Interpreting ML Models (Structured Data) • Using Interpretable Models • Model Feature Importance • Partial Dependence Plots • Individual Conditional Expectation Plots • Local Interpretable Model-Agnostic Explanations • Shapley Values and SHapley Additive Explanations
18. Demo bit.ly/Explainable-AI-A-New-Paradigm-for-Transparency-in-AI
19. Contact Me LinkedIn linkedin.com/in/shadabhussain96/ Website shadabhussain.com Twitter twitter.com/shadabhusain786
20. THANK YOU!