Pythonaifiles

it is the file that i hold my python codes from ai lectures/courses that i take.

Linear classifier to do list

• Split data to test and validation
• Determine attributes
• Create attributes array (add 1 attribute at first add later one by one)
• Do regression(Hinge loss,binarical classification)
• Create test function
• Test with validation and train

Attributes

• Gender(it effects a lot)
• PyClass(it effects a lot too)
• Age is ''
• Has cabin
• Ticket name(it may affect)
• Has multiple cabins
• Fare is between 0-10

Dev Cycle(From Notes)

• Split data into train/val/test
• Look data to get intuition
• Repeat
  • Implement feature/tune hyperparameters
  • Run Learning alg
  • Sanity check train and val error rates weights
  • Look at errors to brainstorm improvements
• Run on test set to get final error rates(ending)

Kaggle da nasil yapiyolar

• Kaggle da bence birseyde hata yapiyorlar featurelari olusturup sonra test ediyorlar test ederke featurelari eklemeleri lazim bence.Sonra feature ekledikten sonra yaptiklari yanlis tahminlere de bakmiyorlar.
• Datanin ai da kullanilacak kisimlarini birakiyorlar.
• Datayi genelde gorsellestiriyorlar
• Sonra direk datayi atip regressyonu yapiyorlar.
• Yasin normal dagilimda su yuzdesinde felan diye de ayirabiliolar.

Dev cycle for kaggle

• Split data into train/val/test
• Look data to get intuition(Use graphs find correlations)
• Repeat
  • Implement feature/tune hyperparameters
  • Run Learning alg
  • Sanity check train and val error rates weights(Use graphs find correlations)
  • Look at errors to brainstorm improvements
• Run on test set to get final error rates(ending)

Dersler bittikten sonra

-K means++ bak -Artik not yok 18/08