AI Roadmaps
← All roadmaps
IntermediateRoadmap 37 · 14 steps · 4h 43m of video

Explainable ML: SHAP, LIME & Model Debugging

Explain what your models do and why: glass-box models, feature importance, PDPs, LIME, SHAP and its limits, counterfactuals, saliency maps, and error analysis.

Before this: Machine Learning. Deep Learning for the Grad-CAM and integrated gradients steps.

  1. 7 min
    Step 1

    What explainable AI is

    Why we need explanations and the main families of methods.

    Explainable AI explained! | #1 Introduction

    DeepFindr

  2. 7 min
    Step 2

    Interpretable vs. explainable models

    Models you can read directly versus explanations added after training.

    Interpretable vs Explainable Machine Learning

    A Data Odyssey

  3. 12 min
    Step 3

    Glass-box models: Explainable Boosting Machines

    Microsoft Research on additive models that stay readable yet accurate.

    The Science Behind InterpretML: Explainable Boosting Machine

    Microsoft Developer

    interpretml/interpret★ 7k
  4. 13 min
    Step 4

    Permutation feature importance

    Shuffle one feature, measure the drop in score, coded from scratch.

    Permutation Feature Importance from Scratch | Explanation & Python Code

    A Data Odyssey

  5. 12 min
    Step 5

    Partial dependence and ICE plots

    See how predictions change as one feature moves, on average and per row.

    Partial Dependence (PDPs) and Individual Conditional Expectation (ICE) Plots | Intuition and Math

    A Data Odyssey

  6. 14 min
    Step 6

    LIME

    Fit a simple local model around one prediction to approximate the black box.

    Explainable AI explained! | #3 LIME

    DeepFindr

    marcotcr/lime★ 12.2k
  7. 15 min
    Step 7

    Shapley values

    The game theory idea behind SHAP: average each feature's marginal contribution.

    Shapley Values : Data Science Concepts

    ritvikmath

  8. 16 min
    Step 8

    SHAP in Python

    Waterfall, force, beeswarm, and dependence plots on a real model.

    SHAP with Python (Code and Explanations)

    A Data Odyssey

    shap/shap★ 25.9k
  9. 7 min
    Step 9

    The limits of SHAP

    Correlated features, baselines, and why SHAP values are not causal effects.

    4 Significant Limitations of SHAP

    A Data Odyssey

  10. 56 min
    Step 10

    Fooling LIME and SHAP

    Talk showing a biased model can be built to get innocent-looking explanations.

    Hima Lakkaraju: How can we fool LIME and SHAP? Adversarial Attacks on Explanation Methods

    Harvard's CRCS

  11. 17 min
    Step 11

    Counterfactual explanations

    The smallest input change that flips a decision, and its link to adversarial examples.

    Explainable AI explained! | #5 Counterfactual explanations and adversarial attacks

    DeepFindr

    interpretml/DiCE★ 1.5k
  12. 14 min
    Step 12

    Saliency maps with Grad-CAM

    Highlight the image regions a CNN's prediction depends on.

    Grad-CAM Explained | FREE XAI Course | L7 - Gradient-weighted Class Activation Mapping

    A Data Odyssey

    jacobgil/pytorch-grad-cam★ 13k
  13. 16 min
    Step 13

    Integrated gradients and feature attribution

    Stanford lecture on gradient attribution methods and the axioms they satisfy.

    Feature Attribution | Stanford CS224U Natural Language Understanding | Spring 2021

    Stanford Online

    meta-pytorch/captum★ 5.7k
  14. 1h 19m
    Step 14

    Debugging models with error analysis

    Andrew Ng's process for diagnosing bias, variance, and pipeline errors.

    Lecture 12 - Debugging ML Models and Error Analysis | Stanford CS229: Machine Learning (Autumn 2018)

    Stanford Online

Build it

Watching isn't learning. Pick one and put it on GitHub.

Browse all roadmaps →