Coursera Machine Learning Exercise #2

Supervised Learning - Classification Problems (Logistic Regression Model)

---

Prerequisites

GNU Octave 4.0.0+

---

Files Included In This Exercise

Drivers (Main Files)

• ex2.m - Run non-regularized logistic regression model.
• ex2_reg.m - Run regularized logistic regression model.
• submit.m - Submit code to Coursera grader.

Datasets

• ex2data1.txt - Used in ex2.m
• ex2data2.txt - Used in ex2_reg.m

Plotting Functions

• plotData.m - Plot 2D classification data.
• plotDecisionBoundary.m - Plot classifier's decision boundary.

Logistic Regression Functions

• sigmoid.m - Logistic Regression Hypothesis Function
• predict.m - Logistic Regression Prediction Function
• costFunction.m - Logistic Regression Cost Function (Unregularized)

Regularization Functions

• mapFeature.m - Function to generate polynomial features, thereby necessitating regularization.
• costFunctionReg.m - Logistic Regression Cost Function (Regularized)

---

Essential Concepts

Terminology

• m = number of training examples
• n = number of features
• = hypothesis function weights; parameters
• x = input
• y = actual output
• = prediction; output of hypothesis function
• = input for training example i, feature j
• = output for training example i

Logistic Regression Hypothesis Function - Sigmoid

Logistic Regression Cost Function (Unregularized)

Logistic Regression Cost Function (Regularized)

Gradient Descent (Unregularized)

For j = [0, n]:

Gradient Descent (Regularized)

For j = 0:

For j = [1,n]:

---

References

• Coursera Machine Learning Exercise #2 Instructions
• GNU Octave Documentation
• Matlab Documentation