コンピュータグラフィックス最適化

授業情報

  • 講師: シモセラ エドガー
  • 日程: 2020年04月〜2020年07月
  • 曜日時限: 火曜日 4時限(14:45〜16:15)
  • 教室: online
  • メール: ess@waseda.jp

オーバービュー

この授業のスライドと講義は主に英語ですが、必要に応じて日本語でも説明します。または、日本語の講義資料もあります。
The class slides, materials, and explanations are primarily in English, however, additional explanations will be given in Japanese as needed.
Due to the effect of the ongoing corona virus, the course has been modified to fit a 12-week schedule and the evaluation has been changed to focus on the final project as there will be no final exam. Furthermore, video explanations will be provided in an "ondemand"-format (videos viewable anytime) and Q&A will be done using Moodle functionality.

This courses focuses on different optimization approaches and their relationship with computer graphics optimization, such as colorization, texture synthesis, or shape manipulation. The course covers both convex and non-linear optimization, including discussion on recent developments such as deep learning. Each lecture will attempt to explain the theoretical foundations of a particular algorithm, and then explain real research examples based on the explained techniques.

The course will be evaluated mainly based on a large project and a final exam. The project will be done in small groups and consist of putting into practice the techniques learned in the class.

The programming assignments and project will be done in python using Jupyter notebooks. While the basics of python will be taught in this class, it is highly recommended that students complement this with self-study using the additional resources.

準備

The following are slides that are meant to refresh basic concepts that are used throughout the lessons. Students are advised to review the following concepts before starting the class.

  1. Python Crash Course 資料 A Whirlwind Tour of Python
  2. Mathematics Refresher 資料 CO Appendix A

授業計画

  1. Overview, Introduction, Least-Squares 資料 CO Chapter 1 NO Chapter 1
  2. Line Search 資料 NO Chapter 2+3
  3. Convex Optimization I: Definitions 資料 CO Chapter 2+3+4
  4. Convex Optimization II: Duality 資料 CO Chapter 5
  5. Unconstrained Minimization 資料 CO Chapter 9 NO Chapter 2
  6. Newton and Quasi-Newton Methods 資料 NO Chapter 6
  7. Derivatives 資料 NO Chapter 8
  8. Linear Programming: Interior-point Methods 資料 CO Chapter 11 NO Chapter 14
  9. Metaheuristics 資料 EM
  10. Deep Learning I: Data, Models 資料
  11. Deep Learning II: Advanced Techniques 資料
  12. Group Work Presentation 資料

教科書

  • CO Stephen Boyd, Convex Optimization, Cambridge University Press, 2004. ISBN: 978-0521833783 website PDF
  • NO Jorge Nocedal and Stephen Wright, Numerical Optimization (2nd Edition), Springer, 2006. ISBN: 978-0387303031
  • EM Sean Luke. Essentials of Metaheuristics. lulu.com, 2013. ISBN: 978-1300549628 website
  • Additional Resources
    • Aharon Ben-Tal and Arkadi Nemirovski, Lectures on Modern Convex Optimization PDF
    • Sébastian Bubeck, Convex Optimization: Algorithms and Complexity PDF
    • Jake VanderPlas, A Whirlwind Tour of Python, O’Reilly Media, 2016. ISBN: 978-1492037859 website
    • Jake VanderPlas, Python Data Science Handbook, O’Reilly Media, 2016. ISBN: 978-1491912058 website
    • Charles R. Severance, Python for Everybody: Exploring Data Using Python 3, Createspace Independent Pub, 2016. ISBN: 978-1530051120 website
    • Aston Zhang, Zack C. Lipton, Mu Li, Alex J. Smola, Dive into Deep Learning, 2019. website