-
-
Notifications
You must be signed in to change notification settings - Fork 136
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #340 from OpenMined/feature/machine-learning
Added Naive Baye Support
- Loading branch information
Showing
21 changed files
with
3,222 additions
and
19 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,219 @@ | ||
{ | ||
"cells": [ | ||
{ | ||
"cell_type": "markdown", | ||
"metadata": {}, | ||
"source": [ | ||
"# Implementation of Naive Bayes in PyDP\n", | ||
"\n", | ||
"*Source*: [Differentially Private Na¨ıve Bayes Classification](https://www.researchgate.net/profile/Anirban-Basu/publication/262254729_Differentially_Private_Naive_Bayes_Classification/links/55dfa68208ae2fac4718fdfd/Differentially-Private-Naive-Bayes-Classification.pdf)" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": 1, | ||
"metadata": {}, | ||
"outputs": [], | ||
"source": [ | ||
"# Cite the paper here\n", | ||
"# Mentioned any part of the paper realted to the implementation\n", | ||
"# Some analysis of the data that is being used\n", | ||
"from sklearn import datasets\n", | ||
"from sklearn.model_selection import train_test_split\n", | ||
"\n", | ||
"dataset = datasets.load_iris()\n", | ||
"X_train, X_test, y_train, y_test = train_test_split(dataset.data, dataset.target, test_size=0.2)" | ||
] | ||
}, | ||
{ | ||
"cell_type": "markdown", | ||
"metadata": {}, | ||
"source": [ | ||
"In order to use the Naive Bayes algorithm with PyDP, we only need to import the `GaussianNB` class from the PyDP's package like the following:\n", | ||
"\n", | ||
"`from pydp.ml.naive_bayes import GaussianNB`\n", | ||
"\n", | ||
"The implementaion is inherited from scikiet learn's Naive Bayes class. Some attributes and methods have been modified to support privacy guarentees. \n", | ||
"\n", | ||
"The following parameters can be adjust according to the use of the algorithm:\n", | ||
"\n", | ||
"- `epsilon`: Privacy parameter for the model. (float, default: 1.0)\n", | ||
" \n", | ||
"- `bounds`: Bounds of the data, provided as a tuple of the form (min, max). `min` and `max` can either be scalars, covering the min/max of the entire data, or vectors with one entry per feature. If not provided, the bounds are computed on the data when ``.fit()`` is first called, resulting in a :class:`.PrivacyLeakWarning`. (tuple, optional)\n", | ||
" \n", | ||
"- `priors`: Prior probabilities of the classes. If specified the priors are not adjusted according to the data. (array-like, shape (n_classes,))\n", | ||
"\n", | ||
"- `var_smoothing` Portion of the largest variance of all features that is added to variances for calculation stability. (float, default: 1e-9)\n", | ||
"\n", | ||
"- `probability`. Probability for a geometric distribution from which a sample will be drawn as noise for categorical features. (float, default: 1e-2)\n", | ||
"\n", | ||
"Source codes:\n", | ||
"- [PyDP's Navie Bayes](https://github.com/OpenMined/PyDP/blob/feature/machine-learning/src/pydp/ml/naive_bayes.py)\n", | ||
"- [Geometric Mechanism in PyDP Naive Bayes' implementation](https://github.com/OpenMined/PyDP/blob/feature/machine-learning/src/pydp/ml/mechanisms/geometric.py)\n", | ||
"- [Laplace Mechanism in PyDP Naive Bayes' implementation](https://github.com/OpenMined/PyDP/blob/feature/machine-learning/src/pydp/ml/mechanisms/laplace.py)" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": 2, | ||
"metadata": {}, | ||
"outputs": [ | ||
{ | ||
"data": { | ||
"text/plain": [ | ||
"GaussianNB(accountant=BudgetAccountant(spent_budget=[(1.0, 0)]),\n", | ||
" bounds=(array([4.3, 2. , 1. , 0.1]), array([7.5, 4. , 6. , 2. ])),\n", | ||
" probability=0.002, var_smoothing=0.0001)" | ||
] | ||
}, | ||
"execution_count": 2, | ||
"metadata": {}, | ||
"output_type": "execute_result" | ||
} | ||
], | ||
"source": [ | ||
"import numpy as np\n", | ||
"\n", | ||
"epsilon = 1.0 # Privacy Budger\n", | ||
"\n", | ||
"lower = np.array([4.3, 2. , 1. , 0.1]) # lower bound of each feature's values\n", | ||
"upper = np.array([7.5, 4. , 6. , 2.]) # upper bound of each feature's values\n", | ||
"\n", | ||
"priors = np.array([0.5, 0.5, 0.5]) # priors of each classes\n", | ||
"\n", | ||
"probability = 0.002 # probability for geometric distribution\n", | ||
"\n", | ||
"var_smoothing = 1e-4 # variance smoothing\n", | ||
"\n", | ||
"from pydp.ml.naive_bayes import GaussianNB\n", | ||
"\n", | ||
"clf = GaussianNB(epsilon=epsilon, bounds=(lower, upper),probability=probability, var_smoothing=var_smoothing)\n", | ||
"clf.fit(X_train, y_train)" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": 3, | ||
"metadata": {}, | ||
"outputs": [ | ||
{ | ||
"data": { | ||
"text/plain": [ | ||
"array([2, 0, 0, 2, 2, 0, 2, 2, 2, 2, 2, 2, 0, 2, 2, 2, 0, 0, 2, 2, 1, 0,\n", | ||
" 2, 2, 0, 2, 0, 2, 2, 2])" | ||
] | ||
}, | ||
"execution_count": 3, | ||
"metadata": {}, | ||
"output_type": "execute_result" | ||
} | ||
], | ||
"source": [ | ||
"y_pred = clf.predict(X_test)\n", | ||
"y_pred" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": 4, | ||
"metadata": {}, | ||
"outputs": [ | ||
{ | ||
"name": "stdout", | ||
"output_type": "stream", | ||
"text": [ | ||
"Accuracy : 0.8\n" | ||
] | ||
}, | ||
{ | ||
"data": { | ||
"text/plain": [ | ||
"array([[ 9, 0, 0],\n", | ||
" [ 0, 1, 6],\n", | ||
" [ 0, 0, 14]])" | ||
] | ||
}, | ||
"execution_count": 4, | ||
"metadata": {}, | ||
"output_type": "execute_result" | ||
} | ||
], | ||
"source": [ | ||
"from sklearn.metrics import confusion_matrix\n", | ||
"from sklearn.metrics import accuracy_score \n", | ||
"\n", | ||
"cm = confusion_matrix(y_test, y_pred)\n", | ||
"print (\"Accuracy : \", accuracy_score(y_test, y_pred))\n", | ||
"cm" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": 5, | ||
"metadata": {}, | ||
"outputs": [ | ||
{ | ||
"name": "stdout", | ||
"output_type": "stream", | ||
"text": [ | ||
"Accuracy : 0.9666666666666667\n" | ||
] | ||
}, | ||
{ | ||
"data": { | ||
"text/plain": [ | ||
"array([[ 9, 0, 0],\n", | ||
" [ 0, 7, 0],\n", | ||
" [ 0, 1, 13]])" | ||
] | ||
}, | ||
"execution_count": 5, | ||
"metadata": {}, | ||
"output_type": "execute_result" | ||
} | ||
], | ||
"source": [ | ||
"# Result from sklearn's version of GaussianNB\n", | ||
"\n", | ||
"from sklearn.naive_bayes import GaussianNB\n", | ||
"\n", | ||
"classifier = GaussianNB()\n", | ||
"classifier.fit(X_train, y_train)\n", | ||
"\n", | ||
"y_pred = classifier.predict(X_test)\n", | ||
"\n", | ||
"cm = confusion_matrix(y_test, y_pred)\n", | ||
"print (\"Accuracy : \", accuracy_score(y_test, y_pred))\n", | ||
"cm" | ||
] | ||
}, | ||
{ | ||
"cell_type": "code", | ||
"execution_count": null, | ||
"metadata": {}, | ||
"outputs": [], | ||
"source": [] | ||
} | ||
], | ||
"metadata": { | ||
"kernelspec": { | ||
"display_name": "Python 3", | ||
"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": "3.6.9" | ||
} | ||
}, | ||
"nbformat": 4, | ||
"nbformat_minor": 4 | ||
} |
Oops, something went wrong.