34
Feature Engineering Studio October 14, 2013

Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Embed Size (px)

Citation preview

Page 1: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Feature Engineering Studio

October 14, 2013

Page 2: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Iterative Feature Refinement

Page 3: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Who here

• Used the Excel Equation Solver• Did not use the Excel Equation Solver

Page 4: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Excel Equation Solver Users

• Sort yourself by the town you were born in (in Roman letters)

Page 5: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Excel Equation Solver Users

• Pick one feature

• What feature did you improve?• What parameter did you adjust?• What was the original and final value?• How big an improvement did you obtain?• Did this process change the meaning of the

feature?

Page 6: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Everyone Else

• Sort yourself by the town you were born in (in Roman letters)

Page 7: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Everyone Else

• Pick one feature

• What feature did you improve?• What parameter did you adjust?• What values did you try?• How big an improvement did you obtain?• Did this process change the meaning of the

feature?

Page 8: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Comments? Questions? Thoughts?

Page 9: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Question

• Is the excel equation solver likely to change the meaning of the feature more than hand processes?

Page 10: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Question

• Is it a good thing or a bad thing, when your feature changes meaning due to refinement?

Page 11: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Feature Parameter Space

• I need a volunteer who had a final best feature that was quite different from their original feature

Page 12: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

One interesting exercise

• I need a volunteer who had a final best feature that was quite different from their original feature

• Please bring up your laptop or a flash drive with your data set

Page 13: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Making…

• A line graph

• X axis – parameter value• Y axis – model goodness

Page 14: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Another volunteer?

• Would anyone else like to look at their feature this way?

• Multiple volunteers are welcome

Page 15: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

What does it mean?

Page 16: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Questions? Comments? Thoughts?

Page 17: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Assignment 6

• Feature Adaptation“This One’s For Nikolai Ivonavich Lobachevsky”

Page 18: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Nikolai Ivonovich Lobachevsky(by Tom Lehrer)

“I will never forget the day I first met the great Lobachevsky.In one word he told me the secret of success in mathematics:”

Page 19: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Nikolai Ivonovich Lobachevsky(by Tom Lehrer)

“I will never forget the day I first met the great Lobachevsky.In one word he told me the secret of success in mathematics:Plagiarize!”

Page 20: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Nikolai Ivonovich Lobachevsky(by Tom Lehrer)

“I will never forget the day I first met the great Lobachevsky.In one word he told me the secret of success in mathematics:Plagiarize!”

“Only be sure to always call it – please – research.”

Page 21: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

To be clear…

• Plagiarism: bad

Page 22: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

To be clear…

• Plagiarism: bad

• Borrowing ideas (and citing them): good

Page 23: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

To be clear…

• Plagiarism: bad

• Borrowing ideas (and citing them): good

• We all clear?

Page 24: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Assignment 6

• You need to find a previous paper that uses one or more features that can be potentially translated to your current analysis

• Find the paper• Try at least one feature in your own data set

Page 25: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Assignment 6

• You need to create a 5-minute presentation– Time yourself to make sure it only runs 5 minutes– To be presented in class next Monday

Page 26: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Assignment 6

• This presentation should discuss– The paper you drew inspiration from– Give a full citation and show us pictures of as many

authors as you can find– The construct being predicted in this paper– The context/data set in this paper– The feature you decided to adapt– The feature you ended up creating– Differences between the original paper’s feature and

your feature– The goodness of your feature in your data set

Page 27: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

What if you can’t find a paper?

Page 28: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

What if you can’t find a paper?

• You can find a paper

Page 29: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

What if you can’t find a paper?

• You can find a paper• Try google scholar

Page 30: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

What if you can’t find a paper?

• You can find a paper• Try google scholar• Email me – but only after you have spent at

least 2 hours searching the web

Page 31: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Questions? Comments?

Page 32: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Upcoming Classes

• 10/16 No special session today

• 10/21 Feature Adaptation– Assignment 6 due

• 10/23 Special Session on Building Prediction Models

Page 33: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Upcoming Classes

• 10/28 Feature Reuse

• 10/30 No special session today

Page 34: Feature Engineering Studio October 14, 2013. Iterative Feature Refinement

Thank you!