本課程以概念講析、理論說明與動手操作的方式,帶領學生踏入AI人工智慧的世界,以具備人工智慧基礎知識為目標。學生不需具備艱深的數學與演算法等專業知識,只要根據本課程的進度循序漸進的學習,就可以了解人工智慧、機器學習、與深度學習的基本概念,且能具備學習新科技與時俱進的能力。
This course takes online-merge-offline methodology introducing business applications to develop students' concepts of artificial intelligence, deep learning, and machine learning, with hands on practices. After activating students' motivations on learning AI, classical algorithms will be introduced. Students will study and design the relevant applications with strong AI theory.
先修科目Prerequisites
無
None
教學方式Teaching Methods
講課
Lecturing
習題練習、書面報告
Exercise, Written report
評量方式Assessment
專題報告
Project report
參考書目Reference
1. Artificial Intelligence: A Modern Approach, Stuart Russell and Peter Norvig. Prentice Hall, 2009 (3rd edition) …1994(1st)…2003(2nd)
2. Nvidia Deep Learning Institute, https://www.nvidia.com/en-us/deep-learning-ai/education/
3. 人工智慧導論,鴻海教育基金會,全華出版社,2019
4. 其他補充資料
教學進度Course Schedule
2020/10/09 人工智慧、機器學習、深度學習介紹
1. 人工智慧、機器學習與深度學習的關係
2. 機器學習介紹
3. 機器學習分類
4. 深度學習介紹
5. 總結與分析比較
Introduction to artificial intelligence, machine learning and deep learning
1. What is artificial intelligence
2. Introduction to machine learning
3. The types of machine learning algorithms
4. Introduction to deep learning
5. Summary
江玠峰(Chiang, Chieh-Feng) 2020/10/16 認識Nvidia提供之線上運算資源與環境
1. 深度學習(Deep Neural Network)介紹
2. 圖形化運算單元(GPU)介紹
3. 大數據(Big Data)介紹
4. 人工智慧運算環境介紹
Along with the course “Fundamentals of Deep Learning for Computer Vision” by Nvidia, students will learn implementing common deep learning workflows, such as image classification and object detection.
江玠峰(Chiang, Chieh-Feng) 2020/10/23 使用深度學習建立應用程式
1. 深度學習專案(Project)建立介紹
2. 深度學習環境佈署
Along with the course “Fundamentals of Deep Learning for Computer Vision” by Nvidia, students will learn experimenting with data, training parameters, network structure, and other strategies to increase performance and capability.
江玠峰(Chiang, Chieh-Feng) 2020/10/30 深度學習效能檢測與改進
1. 深度學習成效分類
2. 佈署預先訓練過的網路模組
3. 影像分類的概念
4. 實務操作
Along with the course “Fundamentals of Deep Learning for Computer Vision” by Nvidia, students will learn deploying neural networks to start solving real-world problems.
江玠峰(Chiang, Chieh-Feng) 2020/11/06 實務操作演練
1. 訓練深度類神經網路
2. 佈署深度類神經網路於Nvidia環境上
Along with the course “Fundamentals of Deep Learning for Computer Vision” by Nvidia, students will learn measuring and Improving Performance.
江玠峰(Chiang, Chieh-Feng) 2020/11/13 在Windows安裝TensorFlow與Keras
1. 安裝Anaconda教學
2. 命令提示字元操作環境介紹
3. 建立TensorFlow的Anaconda虛擬環境
4. 在Anaconda虛擬環境安裝TensorFlow與Keras
5. 啟動Jupyter Notebook
Install Anaconda, TensorFlow and Keras on the windows system
1. Introduction to command prompt
2. Establish Anaconda's virtual environment
3. Install TensorFlow and Keras
4. Start Jupyter Notebook
江玠峰(Chiang, Chieh-Feng) 2020/11/20 Keras MNIST手寫數字辨識資料集介紹
1. 下載MNIST資料教學
2. 查看單筆訓練資料
3. 查看多筆訓練資料images與label
4. 多層感知器模型資料預處理
5. Features(特徵值)資料處理
6. Labels(正確值)資料處理
Introduction to the MNIST dataset
1. Download the MNIST data
2. Check the single training data
3. View the training data and the testing data
4. Data preprocessing
5. Feature Selection
6. Labels
江玠峰(Chiang, Chieh-Feng) 2020/11/27 Keras多元感知器(MLP)辨識手寫數字
1. Keras多元感知器(MLP)辨識Mnist手寫數字介紹
2. 進行資料預處理
3. 建立模型
4. 進行訓練
5. 以測試資料評估模型準確率
6. 進行預測
7. 隱藏層增加為1000個神經元
8. 總結與討論
Introduction to Multilayer perceptron (MLP)
1. Introduction to MLP Model
2. Data preprocessing
3. Modeling
4. Training the network using the training dataset
5. Predicted results
6. Selection of number of nodes in hidden layer
7. Discussion
江玠峰(Chiang, Chieh-Feng) 2020/12/04 期中考
Midterm
江玠峰(Chiang, Chieh-Feng) 2020/12/11 影像分類 (Image Classification)
1. 載入影像資料進訓練環境當中
2. 選擇並訓練網路
3. 測試新資料並逐步提高效能與準確度
Image classification
Train a deep neural network to recognize handwritten digits by:
1. Loading image data to a training environment
2. Choosing and training a network
3. Testing with new data and iterating to improve performance
江玠峰(Chiang, Chieh-Feng) 2020/12/18 物體偵測 (Object Detection)
1. 結合傳統電腦視覺與深度學習
2. 使用深度學習架構Caffe
Object detection
1. Combining traditional computer vision with deep learning
2. Performing minor “brain surgery” on an existing neural network using the deep learning framework Caffe
3. Harnessing the knowledge of the deep learning community by identifying and using a purpose-built network and end-to-end labeled data
江玠峰(Chiang, Chieh-Feng) 2020/12/25 影像分割 (Image Segmentation)
1. 比較影像分割與其他電腦視覺的問題
2. 運用Tensorflow工具如TensorBoard
3. 學習實作有效能的測量工具
Image Segmentation
Learn how to segment MRI images to measure parts of the heart by:
1. Comparing image segmentation with other computer vision problems
2. Experimenting with TensorFlow tools such as TensorBoard and the TensorFlow Python API
3. Learning to implement effective metrics for assessing model performance
江玠峰(Chiang, Chieh-Feng) 2021/01/01 訊號處理 (Signal Processing)
1. 將非影像資料視為影像資料來處理
2. 在Digits中實作一深度學習的流程
3. 提高效能與準確度
Signal processing
1. Treat non-image data as image data
2. Implement a deep learning workflow (load, train, test, adjust) in DIGITS
3. Test performance programmatically and guide performance improvements
江玠峰(Chiang, Chieh-Feng) 2021/01/08 古典人工智慧理論探討-Uninformed search
1. 樹與圖形的搜尋法
2. 廣度搜尋法、深度搜尋法
Uninformed search
1. Search on Tree and Graph
2. Breadth-First Search, Depth-First Search, Uniform-Cost Search
江玠峰(Chiang, Chieh-Feng) 2021/01/15 古典人工智慧理論探討-Informed search
1. 最佳搜尋法
2. 貪婪搜尋法
3. A* 搜尋法及其最佳化
Informed search
1. Best-First Search - Greedy Search, A* Search, Optimality of A*
2. Heuristic - Preformance, Generating Heuristics
江玠峰(Chiang, Chieh-Feng) 2021/01/22 古典人工智慧理論探討-Adversarial search
1. Alpha-Beta 刪除法
Adversarial search
1. Alpha-Beta Pruning
江玠峰(Chiang, Chieh-Feng) 2021/01/29 古典人工智慧理論探討-命題邏輯
1. 推論、分析(Resolution)、CNF(conjunctive normal form)
2. 前向連鎖 (forward chaining) 工作分析法
3. 倒向連鎖 (backward chaining) 工作分析法
Propositional Logic
1. Inference - Simple Knowledge, Resolution and CNF
2. Forward Chaining
3. Backward Chaining
江玠峰(Chiang, Chieh-Feng) 2021/02/05 期末考
Final exam & project report
江玠峰(Chiang, Chieh-Feng)