week 11

Page 1

machine learning pipeline:

segmenting the overall problem into tasks to accomplish the objective

ex., Optical character recognition from Photos:

Image => Character detection => Character segmentation => Character recogniton

Page 2

artificial training example:

have low bias classifier; otherwise adding data sets using artificial data sysntheis wont help;

ceiling analysis :

pipeline:

overall accuracy : 72%

go thru the each step : manually evaluate the next step ; get accuracy

get new improved accuracy of each step ;

you have to focus on the right steps to improve the accuracy;