Created
May 11, 2013 13:28
-
-
Save rishabhjain/5559975 to your computer and use it in GitHub Desktop.
Logistic Regression
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| require 'matrix' | |
| require 'debugger' | |
| def online_gradient_descent(sample,label,weight,learning_rate=0.1) | |
| # debugger | |
| # Expecting Sample and Weight to be a Vector | |
| # Expecting label to be from {1,-1} and learning_rate to be float | |
| # Single epoch, updates weight vector using a single sample | |
| weight -= (-1)*(learning_rate)*label*sample/(1+Math.exp(label*(sample.inner_product(weight)))) | |
| end | |
| def prediction(weight,sample) | |
| exponent = Math.exp(weight.inner_product(sample)) | |
| Math.exp(exponent)/(1+Math.exp(exponent)) | |
| end | |
| def stochastic_gradient_descent(samples,labels,weight=nil,learning_rate=0.1,epochs=5,shuffle=true) | |
| # Expecting samples to be N `cross` M Matrix, N is total number of samples and M is total features. | |
| weight ||= Vector.elements([0]*samples.column_count) #think about how to initialize this properly | |
| 1.upto(epochs).each do | |
| samples = shuffle ? Matrix[*samples.to_a.shuffle] : samples #is it efficient? | |
| # p weight | |
| 0.upto(samples.row_count-1).each do |row_index| | |
| sample = samples.row(row_index) | |
| label = labels[row_index] | |
| # p weight,row_index | |
| weight = online_gradient_descent(sample,label,weight,learning_rate) | |
| end | |
| # p weight | |
| end | |
| weight | |
| end | |
| sample = Matrix[[1,1],[0,1],[0,0],[1,0]] | |
| label = Vector[1,-1,-1,1] | |
| weight = Vector[1,1] | |
| p stochastic_gradient_descent(sample,label) | |
| # Regularization |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment