Time: Wed 9:00-10:30, Fri 10:30-12:00

Location: Engineering Bd 1, Room 509

Textbook: Python Machine Learning, Sebastian Raschka, PACKT Publishing

Lecture Schedule:

• • Oct 27 (Fri), Nov 1 (Wed): no lecture

  • Nov 3 (Fri): regular schedule

  • Nov 8 (Wed): 9:00-10:30, and 13:00-14:30 (both at the classroom, Eng Bd #1, Room 509)

Exams:

• • Midterm: Oct 25 13:00-15:00

    • Location: Enginnering Bd 1, R101

    • Content: all lectures before the exam (closed-book)

    • Results [scores]: Total 50 points, Avg 20.75

  • Final exam: Dec 6, 18:00-20:00

    • Location: Enginnering Bd 1, R101

    • Content: all lectures from the beginning of the semester untill Dec 1st (closed-book)

    • Results [scores]: Total 50 points, Avg 25.43

    • Exam sheet checking: Dec 18~19, 13:00~15:00 Artificial Intelligence Lab (Cluster Bd. R 620)

Mini-Project:

• • MNIST dataset

    • D1: Training set: 60,000 [train-images-idx3-ubyte.gz ] [train-labels-idx1-ubyte.gz ]

    • D2: Test set (1/6 of the entire test set, with labels): 10,000 [test-images-idx3-ubyte.gz ] [test-labels-idx1-ubyte.gz ]

    • D3: Test set (the entire test set, without labels): 60,000 [testall-images-idx3-ubyte.gz ]

  • Apply a machine learning algorithm from the class

  • Do hyperparameter tuning

  • Goal: to achieve the best accuracy on the entire test set, D3.

  • What to submit:

• • 1. A description of your machine learning method (5 pages, A4, PDF, reproducibility)

       • The ML method chosen

       • How the training has been performed (pre-processing, split of validation, CV, etc).

       • Values of hyperparameters, and how they are chosen

    2. Python code to train with the given training data, and to produce label predictions (0~9) of the given teset data:

       • myCLF.py <training_images.gz> <training_labels.gz> <test_images.gz>

         • The filename and the argument format must be the same as above

       • Produce labels to the standard out, one per line: for four test images, produce predictions like

|0

|2

|3

|5

• • 1. A text file containing the prediction results of D3

       • Filename: prediction.txt

• • Submit to the TA via email: jeonghyeonlee@icloud.com

    • Archieve everything to a zip file: <your student ID>.zip

    • Email should be received by Dec 22, 18:00 (no late submission will be accepted)

  • Discussion is encouraged, but you MUST make your own answer (code, report, predictions)

    • Copying others' results will get 0 point

  • Result

    • [link]

    • If you do not submit a prediction.txt, you will receive a 25% penalty.

    • If the number of labels in prediction.txt is less than 60,000, you will also receive a 25% penalty.

    • The person who received the penalty is marked in blue.

    • If neither the report nor the prediction.txt are submitted, the accuracy is scored as zero.

Final grading:

• • Midterm: 40%

  • Final exam: 40%

  • Mini-project: 10%

  • Attendance: 10%

Lecture Notes

• Lecture 01. Introduction [link]

• Lecture 02. Artificial Neural Networks (updated Oct 19) [link]

• Lecture 03. Logistic Regression (update Oct 19) [link]

• Lecture 04. Support Vector Machine (update Oct 19) [link]

• Lecture 05. Decision trees and KNN (update Oct 19) [link]

• Lecture 06. Data pre-processing [link]

• Midterm Summary [link]

• Lecture 07. Dimensionality reduction: PCA [link]

• Lecture 08. Dimensionality reduction: LDA and kernel PCA (update Nov 10) [link]

• Lecture 09: Model Evaluation and Hyperparameter Tuning [link]

• Lecture 10: Ensemble method [link]

• Lecture 11: Sentiment Analysis [link]

• Lecture 12: Regression Analysis (update Dec 1) [link]

• Lecture 13: Clustering Analysis [link]

• Lecture 14: Artificial Neural Network 1 [use the link below]

• Lecture 15: Artificial Neural Network 2 [link]