教學大綱 Syllabus

科目名稱:卷積神經網路與進階神經網路

Course Name: Convolutional Neural Networks and Advanced Neural Networks

修別:選

Type of Credit: Elective

3.0

學分數

Credit(s)

15

預收人數

Number of Students

課程資料Course Details

課程簡介Course Description

The course objective is the introduction of Convolutional Neural Networks and advanced Neural Networks (Large Language Models in this semester) with the infrastructure of PyTorch/TensorFlow and GPUs.

核心能力分析圖 Core Competence Analysis Chart

能力項目說明


    課程目標與學習成效Course Objectives & Learning Outcomes

    The course objective is the introduction of the Convolutional Neural Networks (CNN) and advanced Neural Networks with the infrastructure of PyTorch/TensorFlow and GPUs. Particularly, we will study Large Language Models (LLMs), in this semester. Students will learn from the practice implementation on artificial neural networks techniques, including CNN and some LLMs. At the end of this course, students should gain: (1) the general knowledge on LLMs methodologies, algorithms and implementation, and (2) the hands-on system development experience on programs with PyTorch/TensorFlow and GPU.

    每周課程進度與作業要求 Course Schedule & Requirements

    教學週次Course Week 彈性補充教學週次Flexible Supplemental Instruction Week 彈性補充教學類別Flexible Supplemental Instruction Type

    週次

    課程內容與指定閱讀

    教學活動與課前、課後作業

    學生每週學習投入時間

    (含課堂教學時數)

    1

    Introduction & Overview

    閱讀cs231n.stanford.edu 2024/ Lecture 1 part 1Lecture 1 part 2COS 597G/ lec01、做作業 (install the up-to-date PyTorch or TensorFlow)

    8 hrs/

    2

    Image Classification with Linear Classifiers

    閱讀cs231n.stanford.edu/ Lecture 2、做作業

    3

    Regularization and Optimization

    閱讀cs231n.stanford.edu/ Lecture 3、做作業

    4

    Neural Networks and Backpropagation

    閱讀cs231n.stanford.edu/ Lecture 4、做作業

    5

    Holiday

     

    6

    Image Classification with CNNs

    閱讀cs231n.stanford.edu/ Lecture 5、做作業

    7

    CNN Architectures

    閱讀cs231n.stanford.edu/ Lecture 6、做作業

    8

    Recurrent Neural Networks

    閱讀cs231n.stanford.edu/ Lecture 7、做作業

    8 hrs/

    9

    Attention and Transformers

    閱讀cs231n.stanford.edu/ Lecture 8、做作業

    10

    Self-Supervised Learning

    閱讀cs231n.stanford.edu 2022/ Lecture 14、做作業

    11

    Two in-class presentations & COS 597G / lec02

    閱讀COS 597G/ lec05COS 597G/ lec06COS 597G/ lec02、做作業

    12

    Two in-class presentations & COS 597G / lec03

    閱讀COS 597G/ lec07COS 597G/ lec08COS 597G/ lec02、做作業

    13

    Two in-class presentations & COS 597G / lec03

    閱讀COS 597G/ lec09COS 597G/ lec10COS 597G/ lec02、做作業

    14

    Two in-class presentations & COS 597G / lec03

    閱讀COS 597G/ lec11COS 597G/ lec12COS 597G/ lec02、做作業

    15

    Two in-class presentations & COS 597G / lec04

    閱讀COS 597G/ lec13COS 597G/ lec14COS 597G/ lec02、做作業

    16

    Two in-class presentations & COS 597G / lec04

    閱讀COS 597G/ lec15COS 597G/ lec16COS 597G/ lec02、做作業

    17

    Two in-class presentations & COS 597G / lec04

    閱讀COS 597G/ lec17COS 597G/ lec18COS 597G/ lec02、做作業

    18

    Two in-class presentations & COS 597G / lec04

    閱讀COS 597G/ lec19COS 597G/ lec20COS 597G/ lec02、做作業

     

         

    授課方式Teaching Approach

    40%

    講述 Lecture

    40%

    討論 Discussion

    20%

    小組活動 Group activity

    0%

    數位學習 E-learning

    0%

    其他: Others: 專案

    評量工具與策略、評分標準成效Evaluation Criteria

    1. GRADE DISTRIBUTION:

    Weekly meeting attendance (1% x 17)   17%

    Meeting participation                               18%

    Group peer evaluation form                    25%

    Presentation deliverable                          40%

    Total                                                           100%

     

    2. CONTRIBUTION EVALUATION:

        You are expected to attend each meeting on time with the assigned readings prepared in advance and to contribute to the meeting discussion either by starting the discussion or building on the contribution of others to move the discussion forward. The sharing of your experience and insights is a key part of the leaning process. To build on the contribution of others requires you to listen and to consider the timing of your participation.

        Meaningful meeting participation will be a factor in the determination of your grade. As in all meetings the more you put into a meeting the more you get out of it. We encourage the sharing of ideas with the meeting during meeting discussions. You are of course responsible for all material discussed in meeting even if you are absent. If you miss a meeting you must get notes from someone else in the meeting and you should designate someone to pick up any handouts for you. When you attend meeting you must be on time and remain for the entire meeting.

        The quality and frequency of your contribution will be taken into account in the grading scheme and will include the quality of your responses when cold called. You will be evaluated after every meeting session using the following criteria. Please note that contributions are NOT equivalent to only attending meeting or talking in meeting. The quality of what is said and of one's listening and responding to others are important components of my evaluation.

        Excellent Participation (A): (1) regularly initiates meeting discussions; (2) contributes consistently to meeting discussions; (3) regularly gives indication of substantial knowledge and insights; (4) frequently facilitates others in clarifying and developing their own viewpoints; (5) regularly builds on the thinking of others and integrates that thinking into own contributions to produce a larger synergistic understanding of the issues being discussed.

        Good Participation (B): (1) frequently initiates meeting discussions; (2) contributes consistently to meeting discussions; (3) regularly gives indication of substantial knowledge and insights; (4) occasionally facilitates others in clarifying and developing their own viewpoints.

        Fair Participation (C): (1) occasionally initiates meeting discussions; (2) contributes occasionally to meeting discussions; (3) gives indication of some knowledge and insights; (4) almost never responds constructively to the contribution of others.

        Poor Participation (D): (1) never or almost never initiates meeting discussions; (2) never or almost never contributes to meeting discussions; (3) is late for, does not attend, or is not prepared for 3 or more meetings; (4) actively inhibits or impedes the course of discussion; (5) exhibits defensive behavior such as aggression or withdrawal rather than being thoughtful and considerate of others' ideas.

        Failing Participation (F): (1) never or almost never initiates meeting discussions; (2) never or almost never contributes to meeting discussions; (3) is late for, does not attend, or is not prepared for 6 or more meetings; (4) actively inhibits or impedes the course of discussion; (5) exhibits defensive behavior such as aggression or withdrawal rather than being thoughtful and considerate of others' ideas.

    指定/參考書目Textbook & References

    TBA

    已申請之圖書館指定參考書目 圖書館指定參考書查詢 |相關處理要點

    維護智慧財產權,務必使用正版書籍。 Respect Copyright.

    課程相關連結Course Related Links

    Stanford University CS231n: Convolutional Neural Networks for Visual Recognition
    www.cs.princeton.edu/courses/archive/fall22/cos597G/
    COMP3361 | Tao Yu (余涛) (taoyds.github.io)
    

    課程附件Course Attachments

    課程進行中,使用智慧型手機、平板等隨身設備 To Use Smart Devices During the Class

    Yes

    列印