3 files changed, 76 insertions, 0 deletions
diff --git a/src/logreg_predct.py b/src/logreg_predct.py
new file mode 100644
index 0000000..ae416b0
--- /dev/null
+++ b/src/logreg_predct.py
@@ -0,0 +1,11 @@
+import sys
+
+from model import Model
+
+
+if __name__ == '__main__':
+    if len(sys.argv) != 3:
+        raise 'Usage: {} dataset_path weights_path'.path(*sys.argv[1:])
+    m = Model()
+    m.predict()
+
diff --git a/src/logreg_train.py b/src/logreg_train.py
new file mode 100644
index 0000000..8bc9a25
--- /dev/null
+++ b/src/logreg_train.py
@@ -0,0 +1,11 @@
+import sys
+
+from model import Model
+
+
+if __name__ == '__main__':
+    if len(sys.argv) != 2:
+        raise 'Usage: {} dataset_path'.format(sys.argv[0])
+    m = Model()
+    m.train()
+    # write
diff --git a/src/model.py b/src/model.py
index e69de29..f777b16 100644
--- a/src/model.py
+++ b/src/model.py
@@ -0,0 +1,54 @@
+class Model:
+    def __init__(self, weights_filename='weights'):
+        self.weights_filename = weights_filename
+
+    def train(self, xs, ys, alpha=1, epoch=1000):
+        for _ in range(epoch):
+            theta = theta - alpha * self.gradient(xs, ys)
+
+    def gradient(self, xs, ys):
+        return np.array([self.partial(xs, ys, i) for i in range(len(self.theta))])
+
+    def partial(self, xs, ys, theta_j):
+        total = 0
+        for x_i, y_i in zip(xs, ys):
+            temp = self.hypothesis(x_i) - y_i
+            if theta_j != 0:
+                temp *= x_i[theta_j - 1]
+                total += temp
+                return total / len(xs)
+
+    def hypothesis(self, x):
+        return 1 if  self._sigmoid(x.dot(self.theta)) >= 0.5 else 0
+
+    def logloss(self, x, y):
+        if y == 1:
+            return -np.ln(self.hypothesis(x))
+        elif y == 0:
+            return -np.ln(1 - self.hypothesis(x))
+        else:
+            raise "y != 1 and y != 0"
+
+    def cost(self, xs, ys):
+        return sum([self.logloss(x, y) for x, y in zip(xs, ys)]) / len(xs)
+
+    def _sigmoid(self, x):
+        return 1 / (1 + np.exp(-x))
+
+    def _normalize(self, x):
+        return (x - x.min()) / (x.max() - x.min())
+
+    def _read_weights(self):
+        try:
+            with open(self.weights_filename, 'r') as file:
+                self.weights = np.array(
+                    [float(s) for s in file.read().strip().split(',')])
+        except IOError:
+            raise 'Couldn\'t read weights file at: {}'.format(self.weights_filename)
+
+    def _write_weights(self):
+        try:
+            with open(self.weights_filename, 'w') as file:
+                file.write(','.join([str(w) for w in self.weights])
+        except IOError:
+            raise 'Couldn\'t write weights file at: {}'.format(self.weights_filename)