Machine Learning in Practice:
Do's and Dont's

Suvash Sedhain

Machine Learning?

A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.Tom Mitchell:@:@:@:@:@ WTH !!!!!!!!!!

Machine Learning Class

Supervised

Unsupervised

Reinforcement

Where can I use it?

Dude you are asking Wrong Question!!!!

Where can't I use it?

Dude this is even harder.ML is everywhere.

While (not satisfied){ Write/Improve prototype Run Test }

How to start

Damm!! My Learning Model sucks :(Its not doing right job.

WHAT TO DO ??

DONT PANIC !!!

Diagnose What's wrong :)

Common problem

Under-Fit

Over-Fit

Complex model Sucks

Model Complexity Vs Error

Detect fitting problem

UnderFitHigh training and test error

OverFitLow training error and high test Error

Good-FitYou are satisfied with result

Things to Care about

Accuracy

Precision

Recall

How can you improve model?

Increase training set

Reduce feature vector

Increase feature vector

Improve Optimization process

Use different algorithm

Machine learning Dont's

Premature Optimization

Misunderstanding of Problem

Reinventing the wheels

Incorrect diagnosis of problem

DEMO

Keep Hacking, Make a difference

Click to edit the title text format

Click to edit the title text format

Click to edit the title text format

???Page ??? (???)08/27/2010, 00:36:15Page /