Explaining Machine Learning Models

Details

In many applications, trust in an AI system will come from its ability to ‘explain itself.’ But when it comes to understanding and explaining the inner workings of an algorithm, one size does not fit all. Different stakeholders require explanations for different purposes and objectives, and explanations must be tailored to their needs. While a regulator will aim to understand the system as a whole and probe into its logic, consumers affected by a specific decision will be interested only in factors impacting their case – for example, in a loan processing application, they will expect an explanation for why the request was denied and want to understand what changes could lead to approval.

AI Explainability 360 (AIX360) is an open source toolkit that includes algorithms that span the different dimensions of ways of explaining along with proxy explainability metrics.

In this workshop you will explore different kinds of explanations suited to different users. You will learn:

how to build several machine learning models
how to evaluate these models and their output

Getting Started with Jupyter Notebooks

Jupyter notebooks are an open-source web application that allows you to create and share documents that contain live code, equations, visualizations and explanatory text.

In this workshop we will use IBM Watson Studio to run a notebook. For this you will need an IBM Cloud account. The following steps will show you how sign up and get started. When you have the notebook up and running we will go through the notebook.

IBM Cloud

Go to your IBM Cloud account or sign up

When you have used Watson Studio before, you can skip the next few steps. Go straight to the Resources list and click on Watson Studio and then click on Get Started. This will open a new Tab with Cloud Pak For Data that includes Watson Studio.
When you have just created an account, click Create Resource at the top of the Resources page. You can find the resources under the hamburger menu at the top left:

Search for Watson Studio and click on the tile:

Select the Lite plan and click Create.
Go back to the Resources list and click on your Watson Studio service and then click Get Started. This will open a new Tab with Cloud Pak For Data that includes Watson Studio.

IBM Watson Studio

1. Create a new Project

You should now be in Watson Studio.
Create a new project by clicking on Get Started and New Project, or Create Project
Give your Project a name.

If you have used Watson Studio before you can select an Object Storage from the drop-down menu
If this is your first time using Watson Studio you have to create Object Storagefor free that is used to store the notebooks and data. Follow the instructions and do not forget to click refresh when returning to the Project page.

click Create.

2. Load and run a notebook

Within the new project now add a new notebook. Click Add to project and choose Notebook:

Choose new notebook From URL. Give your notebook a name and copy the URL https://github.com/IBMDeveloperUK/AIX360-workshop/blob/master/notebooks/aix360-workshop.ipynb
Select the Default Python 3.6 XS enviroment and click Create Notebook.
The notebook will load.

You are now ready to follow along with the workshop in the notebook!

Optional: Anaconda local install

Optional local install on Mac:

Install Anaconda

Open terminal and create a new environment:

> conda create --name aix360 python=3.7

> conda activate aix360

Just in case: > conda env remove --name aix360

Add kernel to Jupyter notebooks:

> python -m ipykernel install --user --name aix360 --display-name "Python37 (aix360)"

Start notebooks: > jupyter notebook

IBMDeveloperUK / AIX360-Introduction