SHAP Toolbox

(aiverify.stock.shap-toolbox) [source]

Description

This plugin explains how your features affect your overall predictions by using Shapley Values.

Plugin Content

• Algorithms

Name	Description
SHAP Toolbox	SHAP (SHapley Additive exPlanations) is a game theoretic approach to explain the output of any machine learning model.

• Widgets

Name	Description
Introduction	To provide an introduction to SHAP
Understanding Bar Chart	To guide your users on reading the generated bar chart
Bar Chart (Summary)	To generate bar chart, interpretation and recommendations for explainability testing
Bar Chart (Technical)	To display the average SHAP values using a bar chart
Recommendations	To provide a recommendation for explainability testing

Using the Plugin in AI Verify

Data Preparation

• Tabular dataset (Tutorial for Preparation)

Algorithm User Input(s)

Input Field	Description	Type
Type of explainability	Options: [global (default), local]. Global explainability explains overall dataset. Local explainability explains a random data point.	string
Path of the background data	Background data path	string
Size of the background	Background samples (eg. 25)	int
Size of the test dataset	Data Samples (eg. 25)	int

Sample use of the widgets

More details

Algorithm input schema

{
    "title": "Algorithm Plugin Input Arguments",
    "description": "A schema for algorithm plugin input arguments",
    "type": "object",
    "required": [
        "explain_type",
        "background_path",
        "background_samples",
        "data_samples"
    ],
    "properties": {
        "explain_type": {
            "title": "Type of Explainability",
            "description": "Options: [global (default), local]. Global explainability explains overall dataset. Local explinability explains a random data point.",
            "type": "string",
            "default": "global",
            "enum": [
                "global",
                "local"
            ]
        },
        "background_path": {
            "title": "Path of the Background Path",
            "description": "Background data path",
            "type": "string",
            "ui:widget": "selectDataset"
        },
        "background_samples": {
            "title": "Size of the Background",
            "description": "Background Samples (e.g. 25)",
            "type": "number"
        },
        "data_samples": {
            "title": "Size of the Test Dataset",
            "description": "Data Samples (e.g. 25)",
            "type": "number"
        }
    }
}

Algorithm output schema

{
    "title": "Algorithm Plugin Output Arguments",
    "description": "A schema for algorithm plugin output arguments",
    "type": "object",
    "required": [
        "feature_names",
        "results"
    ],
    "properties": {
        "feature_names": {
            "type": "array",
            "description": "Array of feature names",
            "minItems": 1,
            "items": {
                "type": "string"
            }
        },
        "results": {
            "description": "Matrix of feature values (# feature names)",
            "type": "object",
            "required": [
                "num_local_classes",
                "local",
                "single_explainer_values",
                "single_shap_values",
                "global_shap_values",
                "global_samples",
                "num_global_classes",
                "global"
            ],
            "properties": {
                "num_local_classes": {
                    "description": "Number of local classes",
                    "type": "number"
                },
                "local": {
                    "description": "# of local classes",
                    "type": "array",
                    "minItems": 1,
                    "items": {
                        "type": "array",
                        "minItems": 1,
                        "items": {
                            "type": "array",
                            "description": "class values",
                            "minItems": 1,
                            "items": {
                                "type": "number"
                            }
                        }
                    }
                },
                "single_explainer_values": {
                    "description": "array of single explainer values",
                    "type": "array",
                    "minItems": 1,
                    "items": {
                        "type": "number"
                    }
                },
                "single_shap_values": {
                    "description": "array of single shap values",
                    "type": "array",
                    "minItems": 1,
                    "items": {
                        "type": "array",
                        "description": "class values",
                        "minItems": 1,
                        "items": {
                            "type": "number"
                        }
                    }
                },
                "global_shap_values": {
                    "description": "global shap values",
                    "type": "array",
                    "items": {
                        "type": "array",
                        "description": "Matrix of SHAP values (# samples x # features)",
                        "minItems": 1,
                        "items": {
                            "type": "array",
                            "description": "Array of SHAP values for each feature",
                            "minItems": 1,
                            "items": {
                                "type": "number"
                            }
                        }
                    }
                },
                "global_samples": {
                    "description": "Matrix of feature values (# samples x # features)",
                    "type": "array",
                    "items": {
                        "type": "array",
                        "description": "Array of sample values for each feature",
                        "minItems": 1,
                        "items": {
                            "type": "number"
                        }
                    }
                },
                "num_global_classes": {
                    "description": "Number of global classes",
                    "type": "number"
                },
                "global": {
                    "description": "# of global classes",
                    "type": "array",
                    "items": {
                        "type": "array",
                        "minItems": 1,
                        "items": {
                            "type": "number"
                        }
                    }
                }
            }
        }
    }
}