setup

.gitignore

@@ -0,0 +1,5 @@
+.idea/
+venv
+publish
+dist
+sefr/__pycache__

LICENSE.txt

@@ -0,0 +1,17 @@
+MIT License
+Copyright (c) 2018 YOUR NAME
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

MANIFEST

@@ -0,0 +1,4 @@
+# file GENERATED by distutils, do NOT edit
+setup.py
+sefr/__init__.py
+sefr/sefr.py

README.md

@@ -0,0 +1,6 @@
+# SEFR: A Fast Linear-Time Classifier for Ultra-Low Power Devices
+
+A Python package for the paper [SEFR: A Fast Linear-Time Classifier for Ultra-Low Power Devices](https://arxiv.org/abs/2006.04620) 
+by Hamidreza Keshavarz, Mohammad Saniee Abadeh, Reza Rawassizadeh
+
+Copied from [original implementation](https://github.com/sefr-classifier/sefr)

how_to_use.py

@@ -0,0 +1,5 @@
+from sefr import SEFR
+
+
+if __name__ == '__main__':
+    print(SEFR())

sefr/__init__.py

@@ -0,0 +1 @@
+from sefr.sefr import SEFR

sefr/sefr.py

@@ -0,0 +1,78 @@
+import numpy as np
+
+
+class SEFR:
+    def __init__(self):
+        self.weights = []
+        self.bias = 0
+
+    def fit(self, train_predictors, train_target):
+        """
+        This is used for training the classifier on data.
+        Parameters
+        ----------
+        train_predictors : float, either list or numpy array
+            are the main data in DataFrame
+        train_target : integer, numpy array
+            labels, should consist of 0s and 1s
+        """
+        X = train_predictors
+        if isinstance(train_predictors, list):
+            X = np.array(train_predictors, dtype="float32")
+
+        y = train_target
+        if isinstance(train_target, list):
+            y = np.array(train_target, dtype="int32")
+
+        # pos_labels are those records where the label is positive
+        # neg_labels are those records where the label is negative
+        pos_labels = np.sign(y) == 1
+        neg_labels = np.invert(pos_labels)
+
+        # pos_indices are the data where the labels are positive
+        # neg_indices are the data where the labels are negative
+        pos_indices = X[pos_labels, :]
+        neg_indices = X[neg_labels, :]
+
+        # avg_pos is the average value of each feature where the label is positive
+        # avg_neg is the average value of each feature where the label is negative
+        avg_pos = np.mean(pos_indices, axis=0)  # Eq. 3
+        avg_neg = np.mean(neg_indices, axis=0)  # Eq. 4
+
+        # weights are calculated based on Eq. 3 and Eq. 4
+
+        self.weights = (avg_pos - avg_neg) / (avg_pos + avg_neg)  # Eq. 5
+
+
+        # For each record, a score is calculated. If the record is positive/negative, the score will be added to posscore/negscore
+        sum_scores = np.dot(X, self.weights)  # Eq. 6
+
+        pos_label_count = np.count_nonzero(y)
+        neg_label_count = y.shape[0] - pos_label_count
+
+        # pos_score_avg and neg_score_avg are average values of records scores for positive and negative classes
+        pos_score_avg = np.mean(sum_scores[y == 1])  # Eq. 7
+        neg_score_avg = np.mean(sum_scores[y == 0])  # Eq. 8
+
+        # bias is calculated using a weighted average
+
+        self.bias = (neg_label_count * pos_score_avg + pos_label_count * neg_score_avg) / (neg_label_count + pos_label_count)  # Eq. 9
+
+    def predict(self, test_predictors):
+        """
+        This is for prediction. When the model is trained, it can be applied on the test data.
+        Parameters
+        ----------
+        test_predictors: either list or ndarray, two dimensional
+            the data without labels in
+        Returns
+        ----------
+        predictions in numpy array
+        """
+        X = test_predictors
+        if isinstance(test_predictors, list):
+            X = np.array(test_predictors, dtype="float32")
+
+        temp = np.dot(X, self.weights)
+        preds = np.where(temp <= self.bias, 0 , 1)
+        return preds

setup.py

@@ -0,0 +1,32 @@
+from distutils.core import setup
+
+setup(
+  name = 'sefr',
+  packages = ['sefr'],
+  version = '1.0.1',
+  license='MIT',
+  description = 'A Fast Linear-Time Classifier for Ultra-Low Power Devices',
+  author = 'Simone Salerno',
+  author_email = 'eloquentarduino@gmail.com',
+  url = 'https://github.com/eloquentarduino/sefr',
+  download_url = 'https://github.com/eloquentarduino/sefr/archive/v_101.tar.gz',
+  keywords = [
+    'ML',
+    'microcontrollers',
+    'machine learning'
+  ],
+  install_requires=[
+    'numpy',
+  ],
+  package_data= {},
+  classifiers=[
+    'Development Status :: 5 - Production/Stable',
+    'Intended Audience :: Developers',
+    'Topic :: Software Development',
+    'License :: OSI Approved :: MIT License',
+    'Programming Language :: Python :: 3',
+    'Programming Language :: Python :: 3.4',
+    'Programming Language :: Python :: 3.5',
+    'Programming Language :: Python :: 3.6',
+  ],
+)