理工学研究科Graduate School of Science and Engineering
COS500X3(計算科学 / Computational science 500)深層学習の効率的処理Efficient Processing of Deep Learning
CAP Q HUUCap Quan HUU
授業コードなどClass code etc
学部・研究科Faculty/Graduate school | 理工学研究科Graduate School of Science and Engineering |
添付ファイル名Attached documents | |
年度Year | 2023 |
授業コードClass code | YB039 |
旧授業コードPrevious Class code | |
旧科目名Previous Class title | |
開講時期Term | 秋学期授業/Fall |
曜日・時限Day/Period | 金4/Fri.4 |
科目種別Class Type | |
キャンパスCampus | 小金井 |
教室名称Classroom name | 各学部・研究科等の時間割等で確認 |
配当年次Grade | |
単位数Credit(s) | 2 |
備考(履修条件等)Notes | |
実務経験のある教員による授業科目Class taught by instructors with practical experience | |
カテゴリーCategory | 応用情報工学専攻 |
すべて開くShow all
すべて閉じるHide All
Outline (in English)
This course is an introduction to deep learning and its applications mainly for computer vision. The course includes the fundamental components of modern deep learning systems such as neural network architectures and learning algorithms. Later, some research topics for computer vision such as image classification, object detection, image segmentation, generative models (GANs), attention & transformers, etc., will be covered.
授業で使用する言語Default language used in class
英語 / English
授業の概要と目的(何を学ぶか)Outline and objectives
This course is an introduction to deep learning and its applications mainly for computer vision. The course includes the fundamental components of modern deep learning systems such as neural network architectures and learning algorithms. Later, some research topics for computer vision such as image classification, object detection, image segmentation, generative models (GANs), attention & transformers, etc., will be covered.
到達目標Goal
After this course, students will be able to understand, implement and/or apply deep neural networks to their research. In addition, they will gain a better overview of several important research topics in computer vision.
この授業を履修することで学部等のディプロマポリシーに示されたどの能力を習得することができるか(該当授業科目と学位授与方針に明示された学習成果との関連)Which item of the diploma policy will be obtained by taking this class?
ディプロマポリシーのうち、「DP1」「DP2」「DP3」に関連
授業で使用する言語Default language used in class
英語 / English
授業の進め方と方法Method(s)(学期の途中で変更になる場合には、別途提示します。 /If the Method(s) is changed, we will announce the details of any changes. )
Lectures will be highly interactive. Students will be asked to participate in discussions and ask questions about the lectures.
There will be some deep learning implementations given and students will practice together during the class. The assignment will be in a form of coding practice or questions.
Students will be asked to explore and present their understanding of a topic/project they are interested in deep learning as a final presentation.
アクティブラーニング(グループディスカッション、ディベート等)の実施Active learning in class (Group discussion, Debate.etc.)
あり / Yes
フィールドワーク(学外での実習等)の実施Fieldwork in class
あり / Yes
授業計画Schedule
授業形態/methods of teaching:対面/face to face
※各回の授業形態は予定です。教員の指示に従ってください。
1[対面/face to face]:Course introduction
Course overview, history of computer vision, setting up coding environments
2[対面/face to face]:Basic machine learning review
Linear classification, model representation, loss functions, and basic mathematics review
3[対面/face to face]:Optimization
Stochastic gradient descent (SGD), regularization, weight decay, momentum, etc.
4[対面/face to face]:Neural networks (part I)
Introduction to neurons and the brain, fully-connected networks, forward propagation
5[対面/face to face]:Neural networks (part II)
Backward propagation algorithm
6[対面/face to face]:Training tricks for neural networks
Study several activation functions, normalization, and regularizations techniques
7[対面/face to face]:Convolutional neural networks (CNN)
Introduction to convolutional neural networks, introduce some CNN models for image classification
8[対面/face to face]:Visualizing and understanding models
Feature visualization, understanding the decisions of models, neural style transfer
9[対面/face to face]:Object detection and image segmentation
Study various CNN architectures for object detection and image segmentation
10[オンライン/online]:Generative adversarial networks
Introduction to generative adversarial networks and its applications
11[オンライン/online]:Attention and transformers
Introduction to self-attention and transformers and the applications for computer vision
12[オンライン/online]:Self-supervised learning
Introduction to deep self-supervised learning, image inpainting, contrastive learning methods
13[オンライン/online]:Diffusion Models
Introduction to diffusion models and its applications
14[オンライン/online]:Final presentations
Students present their projects
授業時間外の学習(準備学習・復習・宿題等)Work to be done outside of class (preparation, etc.)
【本授業の準備・復習時間は、各4時間を標準とします。】Homework and final presentation/project
テキスト(教科書)Textbooks
Slides will be printed and distributed
参考書References
- “Deep Learning” by Goodfellow, Bengio, and Courville (available online at https://www.deeplearningbook.org/)
- CS231n: Deep Learning for Computer Vision - Stanford University (available online at http://cs231n.stanford.edu/index.html)
成績評価の方法と基準Grading criteria
Assignment (both in class/homework) 50% + Final presentation/project 50%
学生の意見等からの気づきChanges following student comments
Not applicable
学生が準備すべき機器他Equipment student needs to prepare
- Personal laptop
- Basic knowledge of Machine Learning
- Basic knowledge of Python, and some packages such as Numpy, Pandas, Matplotlib, PyTorch, etc.