kodluyoruz-ds-102

• Data Science 102 Bootcamp that I gave with the association with Kodluyoruz

• I want to give my special thanks to the assistants of this bootcamp for their hard works to help students and their continuous support to the education throughout the bootcamp.

• You can find the lecture videos at:

  🔗 https://www.youtube.com/playlist?list=PLoaCNumrILN8D5rfBtv83g3WspOyI8bhT

syllabus

1. Mathematics Review:

   1. Why does Gradient give us the direction of greatest increase
   2. Linear Algebra

2. Random Forest

   1. Feature Importance
   2. Correlated Variables
   3. Feature Selection
   4. Profiling for speeding up the training
   5. Extrapolation Problem
   6. How to Deal with Extrapolation

3. When not to use random splitting and k fold cross validation ?

4. Naive Bayes

   1. Bayes Formula
   2. Writing Naive Bayes from Scratch

5. Neural Network from Scratch

   1. Why do Neural Networks work?
   2. Why do we normalize input?
   3. Why do we normalize layers?
   4. Writing Matrix Multiplication
   5. Writing Forward and Backward Passes
   6. Training Loop g) Understanding Optimizers

6. Deep Learning for Tabular Data

7. When not to use Softmax?

8. Pytorch Hooks : Wanna see what is going on in your model ?

9. Initialization does matter - A lot-

10. Dataset, DataLoader from scratch

11. Why does Batchnorm work ? No it does not about internal covariate shift

    1. Writing BatchNorm from scratch

12. Writing Learning Rate Finder

13. MixUp augmentation

14. Label Smoothing

15. Wanna train your models fast - try Mixed Precision Training

• Instructor: Engin Deniz Alpman

• Assistants (alphabetically):

  • Ahmet Arif Avcı
  • Burak Bagatarhan
  • Elif Bayındır
  • Fahri Bilici
  • Kubilay Gazioğlu
  • Melis Han
  • Neris Özen
  • Rana Kalkan
  • Uğur Emek

• In creating this course, there is a good deal of knowledge that I acquired from StatQuest and fastai, thank you for creating great contents. You can find their works at links below:

  • StatQuest: https://www.youtube.com/user/joshstarmer
  • fastai: https://www.fast.ai/