clalexni / neural-net

Neural Net

Goal:

• Linear Classification with Sigmoid Regression

Requirements:

• Python 3.6
• the program takes 4 arguments:
  1. training_file
  2. test_file
  3. learning_rate
  4. number of iterations
• run these two command lines

chmod u+x main.py

./main.py [input1] [input2] [input3] [input4]

• for example:

./main.py data/train2.dat data/test2.dat 0.3 400

Notes:

Description:

• fake attribute is explicitly provided in dataset

Update Rule:

• Gradient Descent with logistic regression:
  

w_i = w_i + alpha * (y_k - sigmoid(wx^T)) * x_i * sigmoid'(wx^T)