Skip to content

Latest commit

 

History

History
25 lines (16 loc) · 921 Bytes

README.md

File metadata and controls

25 lines (16 loc) · 921 Bytes

Linear Regression on MNIST

The code is designated for teaching/demonstrating the simplest and most fundamental model of machine learning. The learner should implement the gradient derivation and the update of the bias parameter B of the regression class.

Required packages: numpy, python-mnist, and matplotlib

Questions:

The learning rate is set to 9e-9 in the code. 
Play around with its value and observe the loss value.

Without googling, what is the role of the learning rate based on your observations?

What seems to be a good range of values for the learning rate?

How powerful is linear regression?

Why is the model limited? (Think mathematically and visually (draw a plot))

How can you improve it? (Think about the data and the available features)

A fundamnetal rule has been broken in this code to obtain the learning rate value of 9e-9, 
in regards to data handling, what could it be?