SHAP

How SHAP is impacted by feature dependencies, causal inference and human biasesPhoto by Vicky Sim on UnsplashSHAP is the most popular IML/XAI method. It is a powerful method used to understand how our models make predictions.But don’t let the popularity persuade you.SHAP still has limitations. You need to keep them in mind when making conclusions using the method.We will discuss 3 significant limitations:The first comes from how SHAP values are calculated — we assume feature independenceThe second comes from how we can…

An introduction and a case studyImage by Evan DennisAs explained in my previous post, the framework of SHAP values, widely used for machine learning explainability has unfortunately failed to reflect the casual structure in its results. Researchers have been proposing possible solutions to remove such a limitation. In the article, I will be reviewing one of the proposed alternatives, the Causal SHAP values (CSVs), and give a simple example with detailed computation to illustrate the difference between CSVs and the…

Two examples of the weak points of SHAP values and an overview of possible solutionsSHAP values seem to remove the trade-off between the complexity of machine learning models and the difficulty of interpretation, encouraging researchers and data scientists to design algorithms without worrying about how to understand the prediction given by any black box. But is it always case that SHAP can explain everything property?In this post, we would like to discuss one important weak point of SHAP values by illustrating some…

An overview of how SHAP feature contributions are calculatedPhoto by Klara Kulikova on UnsplashSuppose you (Player 1) and a friend (Player 2) enter a Kaggle contest. You end up winning the first prize of $10,000. Now, you want to split the money fairly. Your friend suggests that you just split it equally. However, your hyperparameter tuning skills are superior. You believe you deserve a larger share as you contributed more to the team. With this in mind, how can you divide the money up fairly?Conveniently, your friend…

Avoid post-processing the SHAP values of categorical featuresPhoto by Andrew Ridley on UnsplashTypically, to model a categorical feature it first needs to be transformed using one-hot encodings. We end up with a binary variable for each category. The problem with this is that each variable will have its own SHAP value. This makes it difficult to see the overall contribution of the original categorical feature. In a previous article, we explored a solution to this. It involved digging into the SHAP values object and…

Are you still confused about the working of SHAP and Shapley values? Let me provide the most simple and intuitive explanation of SHAP and Shapley values in this article.Source — https://github.com/slundberg/shapSHapley Additive exPlanation (SHAP), which is another popular Explainable AI (XAI) framework that can provide model-agnostic local explainability for tabular, image, and text datasets.SHAP is based on Shapley values, which is a concept popularly used in Game Theory. Although the mathematical understanding of…

Adding up SHAP values of categorical features that have been transformed with one-hot encodingsPhoto by Kalineri on UnsplashCategorical features need to be transformed before they can be used in a model. One-hot encoding is a common way to do this: We end up with a binary variable for each category. This is fine until it comes to understanding the model using SHAP. Every binary variable will have its own SHAP value. This makes it difficult to understand the overall contribution of the original categorical feature.A simple…

A stepwise guide for efficiently explaining your models using SHAP.Photo by Pietro Jeng on UnsplashIntroduction to MLlibApache Spark’s Machine Learning Library (MLlib) is designed primarily for scalability and speed by leveraging the Spark runtime for common distributed use cases in supervised learning like classification and regression, unsupervised learning like clustering and collaborative filtering and in other cases like dimensionality reduction. In this article, I cover how we can use SHAP to explain a Gradient…

The ins and outs of a powerful explainable-AI approachPhoto by Héctor J. Rivas on UnsplashThe power and size of machine learning models have grown to new heights in recent years. With greater complexity comes the need for more accountability and transparency—both for the practitioners who build these models and for those who interpret their results.Within the wide field of explainable AI, one approach that has shown great promise (and drawn a lot of attention) is SHAP (from “SHapley Additive exPlanations”); as its…

Explaining text classifiers made easy with Hugging Face Zero Shot Learning with SHAPImage Source — UnsplashExplainable Machine Learning (XML) or Explainable Artificial Intelligence (XAI) is a necessity for all industrial grade Machine Learning (ML) or Artificial Intelligence (AI) systems. Without explainability, ML is always adopted with skepticism, thereby limiting the benefits of using ML for business use-cases.There is a great opportunity of solving data science problems involving text data as text data is available in…