Neuron Object Class in Python (Machine Learning)

Domain : Polynomial/Multiple Regression

Objective:

we are going to learn Regression models with using Training/Validation/Testing Error concepts with Generalization.

Instruction Rules:

IR0 : You are going to build Python Class with the name NeuronR. This class should carry the following items;

a. Class variables:

i. K-Fold (int)

ii. Activation Function (string)

iii. Loss Function (string)

iv. Cost Function (string)

v. Split Type (int)

vi. Training Set, Validation Set, Test Set are all in matrix/vector (array) form.

vii. Split Range (float)

viii. Val_Range (float)

ix. Weights (w0,w1,…..wn)

x. L2 penalty (float)

xi. L1 penalty (float)

xii. Batch Size (int)

xiii. Epoch (int)

b. Class methods

i. BF_SLR(….) method [Brute Force Simple Linear Regression]

ii. PolR_GD(….,L2,L1) method [High Order Polynomial Regression with Gradient Descent]

iii. MLR_GD(….,L2,L1) method [Multiple Linear Regression with Gradient Descent]

iv. CostF(…) method for calculating Cost.

IR1: In Class methods, Brute Force Simple Linear Regression model should try to search only (w0, w1) coefficients The input parameters of this method is search_range1, search_range2.

(Single input “x” is used. If the training set has multiple features like x1, x2, x3…, then we are going to take the first x1 as input)

Ex: BF_SLR(Training_Set, range1, range2)

IR2: In Class methods, PolR_GD method use high order polynomial regression to find solution. PolOrder is used for the complexity degree of the polynomial. (Single input “x” is used. If the training set has multiple features like x1, x2, x3…, then we are going to take the first x1 as input). L2 penalty or L1 penalty is used to regularize the Polynomial solution for generalization. You can use Split_Type as;

- Split_Type=0: Here you may use Training / Test Splitting by defining Split Range

- Split_Type=1: Here you may use Training /Validation/Test Splitting by defining Split Range, Val_Range

- Split_Type=2: Here you may use k-Fold Cross Validation by defining k-Fold.

Ex: PolR_GD ( Training_Set, Val_Set, Test_Set, PolOrder, L2 penalty, L1 penalty, Split_Type)

IR3 : In Class methods, MLR_GD uses all features (x1,x2,…,xN). L2 penalty or L1 penalty is used to regularize the Multiple Linear Regression solution for generalization. You can use Split_Type as;

- Split_Type=0: Here you may use Training / Test Splitting by defining Split Range

- Split_Type=1: Here you may use Training /Validation/Test Splitting by defining Split Range, Val_Range

- Split_Type=2: Here you may use k-Fold Cross Validation by defining k-Fold.

Ex: MLR_GD (Training_Set, Val_Set, Test_Set, L2 penalty, L1 penalty, Split_Type)

IR4 : In Class methods, CostF function is used for calculating Cost Function. This method uses function name as input

parameter. [We can use RSS, MSE, RMSPROP ]

Ex: CostF ( “RSS” )

IR5: The usage of our class can be given such as NeuronR.MLR_GD(……).

Materials:

===========

M1 : Write a report in paper format (not your code included in)

Example paper format can be found in (also in attachment);

M2 : Your code should be written in Python.

M3 : Training and Testing Data can be any data. But for testing purpose you can use the following.

World University Ranking's Data is going to be used as in;

Data: [login to view URL]

---> "world_rank" is out Feature, "quality_of_education" is our prediction.

---> Don't forget, you should exclude some of the (rank,quality) pair for testing

Python Machine Learning (ML) Kiến trúc phần mềm Statistics

ID dự án: #25597388

Về dự án