2018-1 Deep Learning

Time: Wed 3:00-4:30pm, Fri (online lecture)

Location: Cluster Bd. R509 (학연산클러스터 509호)

Textbook: Deep Learning, Goodfellow, Bengio, Courville, MIT Press, 2016

Grading:

    • Homework: 60%

    • Project: 30%

    • Attendance: 10% (5% online + 5% offline)

Project Timeline

    • Brainstorming (May 16)

    • Project Proposal (submission due May 25, 1pm): 2 pages

      • Motivation

      • Problem description (Data, Method, etc)

    • In-class project check (May 30)

    • No class (June 6: memorial day, June 13: election day)

    • Project presentation (June 14): 5 mins per team

      • 16:30 ~ 17:30 @ Eng Bd #1 (1공학관), R.305

    • Final report (due June 22, 1pm): 10 pages

Project data suggetions


  • Submit Proposal

"dongykang@gmail.com"

*The form of E-mail head,title must be like this.

(if you do not keep that form, TA will not read.)

[cse4048][Student ID][name] title


Homeworks

  • HW1 (due April 27 by 12:00pm @ Cluster Bd. #620. No late submission will be accepted)

    • dnn_mnist.py

    • Creating DNN with Tensorflow [Video Lecture 07]

    • Task 2: you can assume that the keys for hidden layers in the dictionary start with the character 'h'

  • HW2 (due June 1, by 12:00pm @ Cluster Bd. #620. No late submission will be accepted)

  • HW3 (due June 15, by 12:00pm @ Cluster Bd. #620. No late submission will be accepted)


Lecture Videos

  • 1. Basic Math (link)

  • 2. Probability (link)

  • 3. MLE (link)

  • 4. Training problem (link)

  • 5. SGD (link)

  • 6. Back-propagation (link)

  • 7. DNN with TF (link)

  • 8. CNN with TF (link)

  • 9. RNN with TF (link)

  • 10. Activation function 1 (link)

  • 11. Activation function 2 (link)

  • 12. Pre-processing (link)

  • 13. Dropout & batch-normalization (link)


Lecture Notes

  • Lecture 01. Introduction [pdf]

  • Lecture 02. Machine learning basics [pdf]

  • Lecture 03. Multi-Layer Perceptron [pdf]

  • Lecture 04. Introduction to Tensorflow [pdf]

  • Lecture 05. Tensorflow: linear regression, logistic regression [code]

  • Lecture 06. Convolutional neural network 1 [pdf]

  • Lecture 07. Convolutional neural network 2 [pdf]

  • No lecture on April 27

    • Make-up lecture on May 2, 1:30~3:00 @ usual classroom (Cluster Bd. PBL Purple)

  • Lecture 08-09. Recurrent neural network [pdf]

  • Lecture 10. Advanced Topics: GPU, Initialization, Hyperparameter Tuning [pdf]

  • Lecture 11. Project discussion

  • Lecture 12. Project discussion

  • Lecture 13. Project check

  • Lecture 14 (June 14) 16:30 ~ 17:30 @ Eng Bd #1, R.305

  • Advanced Topics: Activation functions, Dropout [pdf]