Machine Learning Numset

Working with MNIST dataset of hand written images can be found HERE

Technologies Used:

• Numpy
• Knearest Neighbor algoritthms

Solution

• Converter transorms such data into appropiate binary data

TRAINIMGFNAME = "training-images.bin"
TRAINIMGLABEL = "training-labels.bin"
TESTIMGFNAME  = "testing-images.bin"
TESTLABELS    = "testing-labels.bin"

• Cleaner class takes binary and based on its type p1 p2 p5 transforms it into the same binary format. Look at cleaner.py to understand further.

• Trainer reads and performs knearest, once testing performs a predition which is associated with a label. whichever label its prediction is closest is the overall chosen predition

Run test.py

• to see if the alphabet images were loaded correctly

Evaluation

• Only 83 percent accurate, planning to scale the picture size.