From b2ff54f2d69589903892296a00f022d4c032998a Mon Sep 17 00:00:00 2001
From: Michal Taraszewski <Michal.Taraszewski@se.com>
Date: Sat, 7 Mar 2020 01:02:47 +0100
Subject: [PATCH 1/3] [model] (feat) Initial commit of Lin Reg model

---
 .../LinearRegressionModel-Copy1.py            | 25 ++++++
 .../LinearRegression/LinearRegressionModel.py | 27 ++++++
 .../LinearRegression/Loss.py                  | 17 ++++
 .../LinearRegression/Optimizer-Copy1.py       | 66 +++++++++++++++
 .../LinearRegression/Optimizer.py             | 82 +++++++++++++++++++
 .../LinearRegression/Validator.py             | 41 ++++++++++
 .../LinearRegression/__init__.py              |  0
 .../LinearRegression/untitled-Copy3.txt       |  0
 8 files changed, 258 insertions(+)
 create mode 100644 src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
 create mode 100644 src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
 create mode 100644 src/SupervisedLearning/LinearRegression/Loss.py
 create mode 100644 src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
 create mode 100644 src/SupervisedLearning/LinearRegression/Optimizer.py
 create mode 100644 src/SupervisedLearning/LinearRegression/Validator.py
 create mode 100644 src/SupervisedLearning/LinearRegression/__init__.py
 create mode 100644 src/SupervisedLearning/LinearRegression/untitled-Copy3.txt

diff --git a/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py b/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
new file mode 100644
index 0000000..7484b97
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
@@ -0,0 +1,25 @@
+import numpy as np
+#from src.SupervisedLearning.LinearRegression.Validator import LinRegValidator
+#from src.SupervisedLearning.LinearRegression.utils import *
+
+
+class LinearRegressionModel:
+    def __init__(self, n_features, weights=None, optimizer="GD", measure="MSE"):
+        self.n_features = n_features
+        if weights:
+            self.weights = weights
+        else:
+            self.weights = np.random.rand(self.n_features, 1)
+        self.optimizer = optimizer
+        self.measure = measure
+        #self.validator = LinRegValidator(n_features=n_features)
+        
+    def fit(self, data, target):
+        #self.validator.validate_training(data, target)
+        self.weights = self.optimizer.optimize(data, target, self.measure)
+        
+    def predict(self, data):
+        data = np.c_[np.ones((data.shape[0], 1)), data]
+        return np.dot(data, self.weights)
+        
+    
diff --git a/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py b/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
new file mode 100644
index 0000000..cf9b255
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
@@ -0,0 +1,27 @@
+import numpy as np
+from src.SupervisedLearning.LinearRegression.Optimizer import BatchGradientDecent
+from src.SupervisedLearning.LinearRegression.Loss import MSE
+from src.SupervisedLearning.LinearRegression.Validator import LinRegValidator
+
+
+class LinearRegressionModel:
+    def __init__(self, n_features, optimizer=None, loss=None):
+        self.n_features = n_features
+        if optimizer:
+            self.optimizer = optimizer
+        else:
+            self.optimizer = BatchGradientDecent(learning_rate=1, n_steps=100, save_history=True)
+        if loss:
+            self.loss = loss
+        else:
+            self.loss = MSE()        
+        self.validator = LinRegValidator(n_features=n_features)
+        
+    def fit(self, data, target, weights=None):
+        self.validator.validate_training(data, target)
+        self.weights = self.optimizer.optimize(data, target, loss=self.loss, weights=weights)
+        
+    def predict(self, data):
+        data = np.c_[np.ones((data.shape[0], 1)), data]
+        return np.dot(data, self.weights)
+        
\ No newline at end of file
diff --git a/src/SupervisedLearning/LinearRegression/Loss.py b/src/SupervisedLearning/LinearRegression/Loss.py
new file mode 100644
index 0000000..93e8949
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/Loss.py
@@ -0,0 +1,17 @@
+import numpy as np
+
+class MSE():
+    def __init__(self):
+        pass
+        
+    def _grad(self, forward, weights, data, target):
+        m = data.shape[0]
+        return 2 / m * np.dot(data.T, forward - target)
+
+    
+    def _loss(self, forward, target):
+        return np.square(np.subtract(forward, target)).mean()
+
+    
+    def _forward(self, weights, data):
+        return np.dot(data, weights)
\ No newline at end of file
diff --git a/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py b/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
new file mode 100644
index 0000000..07ebf64
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
@@ -0,0 +1,66 @@
+import numpy as np
+
+class BatchGradientDecent:
+    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
+#    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
+        """
+        This is the default implementation of the Gradient Decent algorithm.
+
+        :param learning_rate: step size
+        :param weights: Initial theta/weights parameters
+        :param n_steps: number of optimization steps
+        :param save_history: flag whether to save gradients and weights, that can be use to
+        debug/analyze the learning progress
+        """
+        self.learning_rate = learning_rate
+
+        self.n_steps = n_steps
+        self.save_history = save_history
+
+        if save_history:
+            self.history = {} # dictionary that keeps track of the previously calculated gradients
+
+#     def optimize(self, weights, data, target):
+    def optimize(self, data, target, measure="MSE", weights=None):
+        if not weights:
+            weights = np.random.rand(data.shape[1] + 1, 1)
+        data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
+        for step in range(self.n_steps):
+            forward = self.__forward(weights, data)
+            loss = self.__loss(forward, target)
+            gradient = self.__grad(weights, forward, data, target, measure)
+            weights = weights - self.learning_rate * gradient
+            if self.save_history:
+                self.__save_history(step, weights, loss, gradient)
+
+        return weights
+
+    def __grad(self, weights, forward, data, target, measure):
+        """
+        TBD
+        """
+        if measure == "MSE":
+            m = data.shape[0]
+#             print("data.shape ", data.shape)
+#             print("data.T.shape ", data.T.shape)
+#             print("weights.shape ", weights.shape)
+#             print("weights.T.shape ", weights.T.shape)
+#             print("target.shape ", target.shape)
+#             print("target.T.shape ", target.T.shape)
+#             print(weights.shape)
+#             print(target.shape)
+            #return 2 / m * np.dot(data.T, np.dot(data.T, weights) - target)
+            return 2 / m * data.T.dot(data.dot(weights) - target)
+
+    
+    def __loss(self, forward, target):
+        return np.square(np.subtract(forward, target)).mean()
+    
+    def __forward(self, weights, data):
+        return np.dot(data, weights)
+
+    def __save_history(self, step, weights, loss, gradient):
+        self.history[step] = {"gradient": gradient,
+                              "loss":loss,
+                              "new_weights": weights
+                              }
diff --git a/src/SupervisedLearning/LinearRegression/Optimizer.py b/src/SupervisedLearning/LinearRegression/Optimizer.py
new file mode 100644
index 0000000..568c6f3
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/Optimizer.py
@@ -0,0 +1,82 @@
+import numpy as np
+from src.SupervisedLearning.LinearRegression.Loss import MSE
+
+class BatchGradientDecent:
+    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
+        """
+        This is the default implementation of the Batch Gradient Decent algorithm.
+        :param learning_rate: step size
+        :param n_steps: number of optimization steps
+        :param save_history: flag whether to save gradients and weights, that can be use to
+        debug/analyze the learning progress
+        """
+        self.learning_rate = learning_rate
+        self.n_steps = n_steps
+        self.save_history = save_history
+        if save_history:
+            self.history = {} # dictionary that keeps track of the previously calculated gradients
+
+
+    def optimize(self, data, target, loss=None, weights=None):
+        if not weights:
+            weights = np.random.rand(data.shape[1] + 1, 1)
+        data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
+        if not loss:
+            loss = MSE()
+        for step in range(self.n_steps):
+            forward = loss._forward(weights, data)
+            loss_value = loss._loss(forward, target)
+            gradient = loss._grad(forward, weights, data, target)
+            weights = weights - self.learning_rate * gradient
+            if self.save_history:
+                self.__save_history(step, weights, loss_value, gradient)
+        return weights
+
+
+    def __save_history(self, step, weights, loss_value, gradient):
+        self.history[step] = {"gradient": gradient,
+                              "loss": loss_value,
+                              "weights": weights
+                              }
+
+class StochasticGradientDecent:
+    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
+        """
+        This is the default implementation of the Stochastic Decent algorithm.
+        :param learning_rate: step size
+        :param n_steps: number of optimization steps
+        :param save_history: flag whether to save gradients and weights, that can be use to
+        debug/analyze the learning progress
+        """
+        self.learning_rate = learning_rate
+        self.n_steps = n_steps
+        self.save_history = save_history
+        if save_history:
+            self.history = {} # dictionary that keeps track of the previously calculated gradients
+
+
+    def optimize(self, data, target, loss=None, weights=None):
+        m = data.shape[0]
+        if not weights:
+            weights = np.random.rand(data.shape[1] + 1, 1)
+        data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
+        if not loss:
+            loss = MSE()
+        for step in range(self.n_steps):
+            random_index = np.random.randint(0, m + 1)
+            X = data[random_index:random_index+1]
+            y = target[random_index:random_index+1]
+            forward = loss._forward(weights, data)
+            loss_value = loss._loss(forward, target)
+            gradient = loss._grad(forward, weights, data, target)
+            weights = weights - self.learning_rate * gradient
+            if self.save_history:
+                self.__save_history(step, weights, loss_value, gradient)
+        return weights
+
+
+    def __save_history(self, step, weights, loss_value, gradient):
+        self.history[step] = {"gradient": gradient,
+                              "loss": loss_value,
+                              "weights": weights
+                              }
diff --git a/src/SupervisedLearning/LinearRegression/Validator.py b/src/SupervisedLearning/LinearRegression/Validator.py
new file mode 100644
index 0000000..bd02ba9
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/Validator.py
@@ -0,0 +1,41 @@
+import numpy as np
+from pandas.api.types import is_numeric_dtype
+
+
+class DataDimError(Exception):
+    def __init__(self, expected, received):
+        message = "".join(map(str, ["Invalid number of features, expected: ", expected, " received: ", received]))
+        super(DataDimError, self).__init__(message)
+
+class DataTargetMissmatch(Exception):
+    def __init__(self, data, target):
+        message = "".join(map(str, ["Number of data examples: ", data ," does not match target: ", target," examples"]))
+        super(DataTargetMissmatch, self).__init__(message)
+
+class DataTypeError(Exception):
+    def __init__(self):
+        message = "".join(map(str,["Invalid type of data, expected numerical."]))
+        super(DataTypeError, self).__init__(message) 
+
+
+class LinRegValidator:
+    def __init__(self, n_features):
+        self.n_features = n_features
+
+    def validate_training(self, data, target):
+        self.__validate_data(data)
+        self.__validate_data_type(data)
+        self.__validate_data_type(target)
+        self.__check_if_data_and_target_match(data, target)
+
+    def __validate_data(self, data):
+        if data.shape[1] != self.n_features:
+            raise DataDimError(data.shape[1], self.n_features)
+
+    def __validate_data_type(self, data):
+        if is_numeric_dtype(data) != True:
+            raise DataTypeError()
+
+    def __check_if_data_and_target_match(self, data, target):
+        if data.shape[0] != target.shape[0]:
+            raise DataTargetMissmatch(data.shape[0], target.shape[0])
\ No newline at end of file
diff --git a/src/SupervisedLearning/LinearRegression/__init__.py b/src/SupervisedLearning/LinearRegression/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/src/SupervisedLearning/LinearRegression/untitled-Copy3.txt b/src/SupervisedLearning/LinearRegression/untitled-Copy3.txt
new file mode 100644
index 0000000..e69de29

From d16c4b71c061ca5e77315a44428d54e7a83c7dd9 Mon Sep 17 00:00:00 2001
From: Michal Taraszewski <Michal.Taraszewski@se.com>
Date: Sat, 7 Mar 2020 01:06:20 +0100
Subject: [PATCH 2/3] [model] (chore) removed trash

---
 .../LinearRegressionModel-Copy1.py            | 25 -------
 .../LinearRegression/Optimizer-Copy1.py       | 66 -------------------
 .../LinearRegression/untitled-Copy3.txt       |  0
 3 files changed, 91 deletions(-)
 delete mode 100644 src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
 delete mode 100644 src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
 delete mode 100644 src/SupervisedLearning/LinearRegression/untitled-Copy3.txt

diff --git a/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py b/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
deleted file mode 100644
index 7484b97..0000000
--- a/src/SupervisedLearning/LinearRegression/LinearRegressionModel-Copy1.py
+++ /dev/null
@@ -1,25 +0,0 @@
-import numpy as np
-#from src.SupervisedLearning.LinearRegression.Validator import LinRegValidator
-#from src.SupervisedLearning.LinearRegression.utils import *
-
-
-class LinearRegressionModel:
-    def __init__(self, n_features, weights=None, optimizer="GD", measure="MSE"):
-        self.n_features = n_features
-        if weights:
-            self.weights = weights
-        else:
-            self.weights = np.random.rand(self.n_features, 1)
-        self.optimizer = optimizer
-        self.measure = measure
-        #self.validator = LinRegValidator(n_features=n_features)
-        
-    def fit(self, data, target):
-        #self.validator.validate_training(data, target)
-        self.weights = self.optimizer.optimize(data, target, self.measure)
-        
-    def predict(self, data):
-        data = np.c_[np.ones((data.shape[0], 1)), data]
-        return np.dot(data, self.weights)
-        
-    
diff --git a/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py b/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
deleted file mode 100644
index 07ebf64..0000000
--- a/src/SupervisedLearning/LinearRegression/Optimizer-Copy1.py
+++ /dev/null
@@ -1,66 +0,0 @@
-import numpy as np
-
-class BatchGradientDecent:
-    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
-#    def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
-        """
-        This is the default implementation of the Gradient Decent algorithm.
-
-        :param learning_rate: step size
-        :param weights: Initial theta/weights parameters
-        :param n_steps: number of optimization steps
-        :param save_history: flag whether to save gradients and weights, that can be use to
-        debug/analyze the learning progress
-        """
-        self.learning_rate = learning_rate
-
-        self.n_steps = n_steps
-        self.save_history = save_history
-
-        if save_history:
-            self.history = {} # dictionary that keeps track of the previously calculated gradients
-
-#     def optimize(self, weights, data, target):
-    def optimize(self, data, target, measure="MSE", weights=None):
-        if not weights:
-            weights = np.random.rand(data.shape[1] + 1, 1)
-        data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
-        for step in range(self.n_steps):
-            forward = self.__forward(weights, data)
-            loss = self.__loss(forward, target)
-            gradient = self.__grad(weights, forward, data, target, measure)
-            weights = weights - self.learning_rate * gradient
-            if self.save_history:
-                self.__save_history(step, weights, loss, gradient)
-
-        return weights
-
-    def __grad(self, weights, forward, data, target, measure):
-        """
-        TBD
-        """
-        if measure == "MSE":
-            m = data.shape[0]
-#             print("data.shape ", data.shape)
-#             print("data.T.shape ", data.T.shape)
-#             print("weights.shape ", weights.shape)
-#             print("weights.T.shape ", weights.T.shape)
-#             print("target.shape ", target.shape)
-#             print("target.T.shape ", target.T.shape)
-#             print(weights.shape)
-#             print(target.shape)
-            #return 2 / m * np.dot(data.T, np.dot(data.T, weights) - target)
-            return 2 / m * data.T.dot(data.dot(weights) - target)
-
-    
-    def __loss(self, forward, target):
-        return np.square(np.subtract(forward, target)).mean()
-    
-    def __forward(self, weights, data):
-        return np.dot(data, weights)
-
-    def __save_history(self, step, weights, loss, gradient):
-        self.history[step] = {"gradient": gradient,
-                              "loss":loss,
-                              "new_weights": weights
-                              }
diff --git a/src/SupervisedLearning/LinearRegression/untitled-Copy3.txt b/src/SupervisedLearning/LinearRegression/untitled-Copy3.txt
deleted file mode 100644
index e69de29..0000000

From efe51dbcb82ee47af0238140e501de4731b5f377 Mon Sep 17 00:00:00 2001
From: Michal Taraszewski <Michal.Taraszewski@se.com>
Date: Sat, 14 Mar 2020 21:52:53 +0100
Subject: [PATCH 3/3] [model] (chore & tests) Refactored Model and Optimizer,
 added tests

---
 .../LinearRegression/LinearRegressionModel.py | 16 ++++---
 .../LinearRegression/Optimizer.py             | 32 +++++++++++---
 .../LinearRegression/tests/__init__.py        |  0
 .../tests/test_LinearRegressionModel.py       | 44 +++++++++++++++++++
 4 files changed, 81 insertions(+), 11 deletions(-)
 create mode 100644 src/SupervisedLearning/LinearRegression/tests/__init__.py
 create mode 100644 src/SupervisedLearning/LinearRegression/tests/test_LinearRegressionModel.py

diff --git a/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py b/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
index cf9b255..36c38f8 100644
--- a/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
+++ b/src/SupervisedLearning/LinearRegression/LinearRegressionModel.py
@@ -1,12 +1,17 @@
+import sys
+import os
+sys.path.append(os.path.join(os.path.dirname(__file__)))
+
 import numpy as np
-from src.SupervisedLearning.LinearRegression.Optimizer import BatchGradientDecent
-from src.SupervisedLearning.LinearRegression.Loss import MSE
-from src.SupervisedLearning.LinearRegression.Validator import LinRegValidator
+from Optimizer import BatchGradientDecent
+from Loss import MSE
+from Validator import LinRegValidator
 
 
 class LinearRegressionModel:
     def __init__(self, n_features, optimizer=None, loss=None):
         self.n_features = n_features
+        self.weights = np.zeros(n_features + 1) # add weight for bias term 
         if optimizer:
             self.optimizer = optimizer
         else:
@@ -17,9 +22,10 @@ def __init__(self, n_features, optimizer=None, loss=None):
             self.loss = MSE()        
         self.validator = LinRegValidator(n_features=n_features)
         
-    def fit(self, data, target, weights=None):
+    def fit(self, data, target):
+        target = target.reshape(-1,)
         self.validator.validate_training(data, target)
-        self.weights = self.optimizer.optimize(data, target, loss=self.loss, weights=weights)
+        self.weights = self.optimizer.optimize(data, target, loss=self.loss, weights=self.weights)
         
     def predict(self, data):
         data = np.c_[np.ones((data.shape[0], 1)), data]
diff --git a/src/SupervisedLearning/LinearRegression/Optimizer.py b/src/SupervisedLearning/LinearRegression/Optimizer.py
index 568c6f3..5025605 100644
--- a/src/SupervisedLearning/LinearRegression/Optimizer.py
+++ b/src/SupervisedLearning/LinearRegression/Optimizer.py
@@ -1,6 +1,18 @@
 import numpy as np
-from src.SupervisedLearning.LinearRegression.Loss import MSE
+from Loss import MSE
 
+class DummyOptim:
+    def __init__(self):
+        """
+        This class is a default implementation of the Optimizer.
+        It is used to test correctness of the LogisticRgressionModel implementation.
+        """
+        pass
+
+    def optimize(self, data, target, loss, weights):
+        return np.arange(weights.shape[0])
+    
+    
 class BatchGradientDecent:
     def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
         """
@@ -18,15 +30,17 @@ def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
 
 
     def optimize(self, data, target, loss=None, weights=None):
-        if not weights:
-            weights = np.random.rand(data.shape[1] + 1, 1)
+        if isinstance(weights, np.ndarray):
+            pass
+        else:
+            weights = np.random.rand(data.shape[1] + 1, 1) # add weight for bias term 
         data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
         if not loss:
             loss = MSE()
         for step in range(self.n_steps):
             forward = loss._forward(weights, data)
             loss_value = loss._loss(forward, target)
-            gradient = loss._grad(forward, weights, data, target)
+            gradient = loss._grad(forward, weights, data, target)   
             weights = weights - self.learning_rate * gradient
             if self.save_history:
                 self.__save_history(step, weights, loss_value, gradient)
@@ -39,6 +53,7 @@ def __save_history(self, step, weights, loss_value, gradient):
                               "weights": weights
                               }
 
+        
 class StochasticGradientDecent:
     def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
         """
@@ -57,9 +72,14 @@ def __init__(self, learning_rate = 1, n_steps = 10, save_history=False):
 
     def optimize(self, data, target, loss=None, weights=None):
         m = data.shape[0]
-        if not weights:
-            weights = np.random.rand(data.shape[1] + 1, 1)
+        if isinstance(weights, np.ndarray):
+            pass
+        else:
+            weights = np.random.rand(data.shape[1] + 1, 1) # add weight for bias term 
         data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
+#         if not weights:
+#             weights = np.random.rand(data.shape[1] + 1, 1)
+#         data = np.c_[np.ones((data.shape[0], 1)), data] # add bias term (x0 = 1) to each instance
         if not loss:
             loss = MSE()
         for step in range(self.n_steps):
diff --git a/src/SupervisedLearning/LinearRegression/tests/__init__.py b/src/SupervisedLearning/LinearRegression/tests/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/src/SupervisedLearning/LinearRegression/tests/test_LinearRegressionModel.py b/src/SupervisedLearning/LinearRegression/tests/test_LinearRegressionModel.py
new file mode 100644
index 0000000..94076c9
--- /dev/null
+++ b/src/SupervisedLearning/LinearRegression/tests/test_LinearRegressionModel.py
@@ -0,0 +1,44 @@
+import sys, os.path
+
+sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
+
+from unittest import TestCase
+from LinearRegressionModel import *
+from Validator import *
+from Optimizer import DummyOptim
+from Loss import MSE
+import numpy as np
+
+
+class TestLinearRegressionModel(TestCase):
+    def setUp(self):
+        self.n_features = 4
+        self.linreg = LinearRegressionModel(self.n_features, optimizer=DummyOptim(), loss=MSE())
+
+    def test_fit_with_dummy_optim(self):
+        n_events = 5
+        data = np.random.randint(10, size=(n_events, self.n_features))
+        target = np.zeros(n_events)
+        self.linreg.fit(data, target)
+
+    def test_fit_should_rise_if_wrong_data_dimensionality(self):
+        n_events = 5
+        wrong_dim = self.n_features-2
+        data = np.random.randint(10, size=(n_events, wrong_dim))
+        target = np.zeros(n_events)
+
+        self.assertRaises(DataDimError, self.linreg.fit, data, target)
+
+    def test_fit_should_rise_if_data_target_not_equal_examples(self):
+        n_events = 5
+        data = np.random.randint(10, size=(n_events, self.n_features))
+        target = np.zeros(n_events-1)
+
+        self.assertRaises(DataTargetMissmatch, self.linreg.fit, data, target)
+
+    def test_predict_with_dummy_optim(self):
+        n_events = 5
+        data = np.random.randint(10, size=(n_events, self.n_features))
+        predictions = self.linreg.predict(data)
+        
+        self.assertEqual(predictions.shape[0], n_events)