115 學年度第 1 學期電機工程學系碩士班王守均教師深度強化學習課程大綱

課程簡介 Course Introduction
開課年度學期 Year / Term	115 學年度第 1 學期
開課班級 Department	電機工程學系碩士班電機系碩博班合選 Master Program, Department of Electrical Engineering
授課方式 Instructional Method	課堂教學、中文
課程電腦代號 Course Reference Number	182010
課程名稱(中文) Course Title(Chinese)	深度強化學習
課程名稱(英文) Course Title(English)	Deep Reinforcement Learning
學分數/時數 Credit Hours	3 / 3
必(選)修 Required / Elective Course	選修 Elective
授課老師 Instructor	王守均
助教 Teaching Assistant
上課時間 Meeting Time	星期三，節次7 Wed, Period 7、8、9
上課教室 Classroom	ZB205
Office Hours

獲獎及補助情形 Awards and Grants

聯合國永續發展目標 (SDGs跨域類別) Sustainable Development Goals, SDGs

SDGs 04.	優質教育：確保有教無類、公平以及高品質的教育，及提倡終身學習 Quality Education：Ensure inclusive and equitable quality education and promote lifelong learning opportunities for all
SDGs 08.	合適的工作及經濟成長：促進包容且永續的經濟成長，讓每個人都有一份好工作 Decent Work and Economic Growth：Promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all

屬於大學社會責任(USR)性質 Courses related to University Social Responsibility（USR)

是

是否運用AI輔助教學 Using Artificial Intelligence (AI) to assist teaching

是

課程目標 Learning Objectives

If deep learning is how machines mimic brain infrastructure, reinforcement learning is how they replicate the mechanics of the mind. Deep learning provides the sensory hardware, allowing algorithms to perceive patterns, recognize images, and process vast oceans of unstructured data. By marrying these two frameworks, deep reinforcement learning (DRL) creates artificial agents capable of navigating complex, unpredictable environments. The deep network processes high-dimensional sensory inputs, acting as the eyes and ears, while the reinforcement learning architecture acts as the goal-driven intellect.

Deep Reinforcement Learning (DRL) is the catalyst that makes practical AI implementation possible. Without DRL, artificial intelligence remains largely confined to theoretical laboratory environments. It is through DRL that AI transitions into a significant, transformative presence in our daily lives. Fortunately, integrating DRL into your research is highly intuitive and does not demand an advanced background in mathematics. To facilitate this journey, we have selected a highly practical textbook. This resource seamlessly blends theoretical insights with real-world application examples, ensuring you grasp core concepts quickly, precisely, and enjoyably.

先修 ( 前置 ) 課程 Prerequisite

Probability, Statistics and Deep Learning background can help but not required as we will conver those required contents in the class.

彈性教學規劃 Flexible Teaching/Planning Schedules

＊本課程實施16+2週彈性教學方案，其中第17、18週之彈性規劃如下：

展演實作

自主學習

課程大綱 Course Syllabus
週次 Week	課程單元大綱 Unit	教學方式 Instructional Method/Style/Teaching Style	參考資料或相關作業 References or Related Materials	評量方式 Grading
1	What is reinforcement learning?	lecture	textbook	you should be able to describe the basic concepts of reinforcement learning
2	Markov decision process	lecture	textbook	understand exploration and exploitation, write program to compute value and policy functions
3	What is Q-learning?	lecture	textbook	Q-learning modeling
4	Deep Network	lecture	textbook	understand Deep Network
5	Q-learning Review	lecture	textbook/HW	expericne repky & target network stability
6	Q-learning project	presentation	share HW results and findings	present homework
7	Policy gradient methods	lecture	textbook	working with OpenAI Gym
8	Actor-critic methods	lecture	textbook	able to implement the methods
9	Evolutionary algorithms	lecture	textbook	understand evaluationary approach, pros and cons
10	Distributional DQN	lecture	textbook	can describe the basic concepts of DDQN
11	Probabilty and Statistics Reivew	lecture	textbook	understand prior and posterior probability
12	Bellman Equation	lecture	textbook	understand Bellman equation
13	DDQN review	lecture/HW assignment	textbook/HW	HW results and findings
14	DDQN project	presentation	share HW results and findings	HW results and findings
15	Curiosity driven exploration	lecture	textbook	understand another type of exploration
16	Multi-agent reinforcement learning	lecture	textbook	understand neightborhood Q-learnin
17	Interpretable reinforcement learning	lecture	textbook	understand attention and relational models
18	Review and Roadmap	lecture	textbook	content review and further development

單一課程對應校能力指標程度 The Degree to Which Single Course Corresponds to School Competence
編號 No.	校核心能力 School Core Competencies	符合程度 Degree of conformity
1	公民力 (Citizen)	4
2	自學力 (Self-learning)	5
3	資訊力 (Information)	5
4	創造力 (Creativity)	5
5	溝通力 (Communication)	5
6	就業力(Employability)	5

單一課程對應系能力指標程度 The Degree to Which Single Course Corresponds to Department Competence
編號 No.	類別 Category	系核心能力 Department Core Competencies	符合程度 Degree of conformity
01	系所	具備專業知識運作及運用之能力	5
02	系所	發掘問題、實驗分析及驗證之能力	5
03	系所	創新思考開發之能力	5
04	系所	協調合作、領導團隊與管理規劃之能力	5
05	系所	中英文寫作與簡報之能力	5
06	系所	強化國際觀與國際交流之能力	5

單一課程對應院能力指標程度 The Degree to Which Single Course Corresponds to College Competence
編號 No.	院核心能力 College Core Competencies	符合程度 Degree of conformity
1	語文能力	5
2	溝通與合作能力	5
3	創新與實踐能力	5
4	專業知能	5

教科書或參考用書 Textbooks or Reference Books

館藏書名 Library Books

深度強化式學習

備註 Remarks

You may download a free copy of the textbook from:
https://vdoc.pub/download/deep-reinforcement-learning-in-action-6ppasj45qt40

※請尊重智慧財產權，不得非法影印教科書※

※ Please respect intellectual property rights and do not illegally photocopy textbooks. ※

教學方法 Teaching Method
教學方法 Teaching Method	百分比 Percentage
問題導向學習	20 %
講述	50 %
專題實作	30 %
總和 Total	100 %

成績評量方式 Grading
評量方式 Grading	百分比 Percentage
個人書面報告	30 %
個人口頭報告	70 %
總和 Total	100 %

成績評量方式補充說明

課程大綱補充資料 Supplementary Material of Course Syllabus