Language

  • Lecture: Korean
  • Material: English / Korean

Textbook

  • Aurélien Géron, Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow (primary)
  • C. Bishop, Pattern Recognition and Machine Learning (supplementary)

Lecture videos (youtube playlist)

Lecture slides

  1. Lecture 01 | Machine Learning Introduction (slide, note)
  2. Lecture 02 | Basic Mathematics (slide, note)
  3. Lecture 03 | Python Practice I
  4. Lecture 04 | Regression (slide, note)
  5. Lecture 05 | AI Department Seminar
  6. Lecture 06 | Clustering I (slide, note)
  7. Lecture 07 | Clustering II (slide, note)
  8. Lecture 08 | Classification I (slide, note)
  9. Lecture 09 | Classification II (slide, note)
  10. Lecture 10 | Python Practice II
  11. Lecture 11 | Support Vector Machine I (slide, note)
  12. Lecture 12 | Support Vector Machine II (slide, note)
  13. Lecture 13 | Decision Tree and Ensemble Learning (slide, note)
  14. Lecture 14 | Mid-Term Practice (note)
  15. Lecture 15 | Dimensional Reduction I (slide, note)
  16. Lecture 16 | Dimensional Reduction II (slide, note)
  17. Lecture 17 | Neural Network and Backpropagation I (slide, note)
  18. Lecture 18 | Neural Network and Backpropagation II (slide, note)
  19. Lecture 19 | Neural Network and Backpropagation III (note)
  20. Lecture 20 | AI Department Seminar
  21. Lecture 21 | Convolutional Neural Network (slide, note)
  22. Lecture 22 | Python Practice III
  23. Lecture 23 | Recurrent Neural Network (slide, note)
  24. Lecture 24 | Autoencoder (slide, note)
  25. Lecture 25 | Final Exam Practice (note)

Correction

  • SVM: Gaussian kernel에서 gamma값이 서로 다르더라도 두 그래프가 교차하지 않습니다. Gaussian distribution과는 다르게 normalize하는 항이 앞에 붙지 않기 때문입니다. 그래서 gamma값이 작을수록 폭이 좁으면서도 항상 더 작은 값으로만 나타납니다. distance가 0일 때 1이고 모두 여기서 교차합니다. gamma값이 작을수록 거리에 대한 영향이 적어집니다.
  • Ensemble Learning: soft voting classification에서 확률의 평균을 구할 때 결과값과 확률을 곱해서 구하지 않습니다. 확률값 자체의 산술평균을 구합니다.
  • Positive definiteness: singular value가 아닌 eigenvalue를 기준으로 모든 값이 0보다 클 때 성립합니다.

Exam