meachine learing using python

Assignment 2A (25 points) Due Date: 10/15 (11:59pm) FRIDAY Description: In this assignment, you are going to write a python program to implement logistics regression for digit recognition. The image size is 28 x 28. Here is one of the training example.

Here are the tasks for this assignment

• Add ones to the first row to the data matrix

• Make the indicator variable associated with class labels

[[0. 1. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] ... [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 1. 0. 1.] [0. 0. 0. ... 0. 0. 0.]]

• Initialize a random weight matrix whose size is the image dimension * num of classes

[[1. 1. 1. ... 1. 1. 1.] [0. 0. 0. ... 0. 0. 0.] [0. 0. 0. ... 0. 0. 0.] ...

• Next, you need write a train method performing the following steps ▪ Write a softmax function to compute

where W is the weight matrix, T is the matrix transpose, and X is the data matrix

▪ Compute the gradient which is X * (P(y | z) – indicator)

▪ Perform gradient descent

o W = W – learning_rate * gradient

## Prob, W, X, indicator and Gradient are matrices def train(): Loop for max_epoch Prob <- softmax(W, X) Gradient <- X * (Prob – indicator) W <- W – learning_rate * Gradient return W

• Save the W to a pickle file To Run: >> python TrainLogit.py Submission: You need to ZIP up the entire folder and submit the ZIP file to blackboard ** If the file is not a ZIP file, it will be 10 points deduction.