wehub-resource-sync
164 lines
4.4 KiB
Plaintext
164 lines
4.4 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"# Migrating to the new \"Explanation\" API\n",
|
|
"\n",
|
|
"This notebook demonstrates some of the changes to the shap API that were introduced in shap `v0.36.0`."
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 1,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"# An example dataset and model\n",
|
|
"import xgboost\n",
|
|
"\n",
|
|
"import shap\n",
|
|
"\n",
|
|
"X, y = shap.datasets.adult(n_points=100)\n",
|
|
"model = xgboost.XGBClassifier().fit(X, y)\n",
|
|
"explainer = shap.TreeExplainer(model, X)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"To summarise the main change in the API:"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 2,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"shap_values = explainer.shap_values(X) # Old style\n",
|
|
"explanation = explainer(X) # New style"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Calculating explanations\n",
|
|
"\n",
|
|
"### Old style\n",
|
|
"\n",
|
|
"In versions of shap before `v0.36.0`, explanations are represented as simple numpy arrays and calculated using the `.shap_values()` method of an explainer:"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 3,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"data": {
|
|
"text/plain": [
|
|
"array([[-0.54854601, 0.01639348, -0.46476041, 0.85896822, -1.36168788,\n",
|
|
" -0.64692199, 0.0254638 , -0.58422904, -0.02344483, 0. ,\n",
|
|
" 0.1224989 , 0.01079906],\n",
|
|
" [-0.83802091, 0.01562196, 0.78349799, -1.10456323, -0.68524691,\n",
|
|
" -0.84828204, 0.03734176, -0.86151311, -0.02564897, 0. ,\n",
|
|
" -0.56183428, 0.00415988]])"
|
|
]
|
|
},
|
|
"execution_count": 3,
|
|
"metadata": {},
|
|
"output_type": "execute_result"
|
|
}
|
|
],
|
|
"source": [
|
|
"shap_values = explainer.shap_values(X)\n",
|
|
"shap_values[:2] # a numpy array"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Similarly, legacy plotting functions like `shap.summary_plot` expected the `shap_values` as a numpy array."
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"### New style\n",
|
|
"\n",
|
|
"As of shap `v0.36.0`, explanations are now represented with the `Explanation` object, and are created by calling the explainer directly as a function:"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 4,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"data": {
|
|
"text/plain": [
|
|
".values =\n",
|
|
"array([[-0.54854601, 0.01639348, -0.46476041, 0.85896822, -1.36168788,\n",
|
|
" -0.64692199, 0.0254638 , -0.58422904, -0.02344483, 0. ,\n",
|
|
" 0.1224989 , 0.01079906],\n",
|
|
" [-0.83802091, 0.01562196, 0.78349799, -1.10456323, -0.68524691,\n",
|
|
" -0.84828204, 0.03734176, -0.86151311, -0.02564897, 0. ,\n",
|
|
" -0.56183428, 0.00415988]])\n",
|
|
"\n",
|
|
".base_values =\n",
|
|
"array([-2.70354599, -2.70354599])\n",
|
|
"\n",
|
|
".data =\n",
|
|
"array([[27., 4., 10., 0., 1., 1., 4., 0., 0., 0., 44., 39.],\n",
|
|
" [27., 4., 13., 4., 10., 0., 4., 0., 0., 0., 40., 39.]])"
|
|
]
|
|
},
|
|
"execution_count": 4,
|
|
"metadata": {},
|
|
"output_type": "execute_result"
|
|
}
|
|
],
|
|
"source": [
|
|
"explanation = explainer(X)\n",
|
|
"explanation[:2] # a shap.Explanation object"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"The `shap.Explanation` object is a much richer representation that includes the shap values (accessible with the `.values` attribute) as well as supporting contextual information such as the background dataset and the feature names.\n",
|
|
"\n",
|
|
"The new-style plotting functions like `shap.plot.bar` and `shap.plots.beeswarm` accept these `Explanation` objections rather than numpy arrays."
|
|
]
|
|
}
|
|
],
|
|
"metadata": {
|
|
"kernelspec": {
|
|
"display_name": "shap",
|
|
"language": "python",
|
|
"name": "python3"
|
|
},
|
|
"language_info": {
|
|
"codemirror_mode": {
|
|
"name": "ipython",
|
|
"version": 3
|
|
},
|
|
"file_extension": ".py",
|
|
"mimetype": "text/x-python",
|
|
"name": "python",
|
|
"nbconvert_exporter": "python",
|
|
"pygments_lexer": "ipython3",
|
|
"version": "3.11.9"
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|