(教育部智慧聯網技術與應用人才培育計畫-107年度磨課師課程發展計畫(國立成功大學 電機系 李順裕 教授))

Python Deep Learning
code(Packt)
Python程式設計學習經典：工程分析x資料處理x專案開發 (碁峰)(作者： 吳翌禎, 黃立政 ISBN：9789864768837)(出版日期：2018/08/23)
林政益 (scott_lin@gotop.com.tw) 電話： 04-2452-7051 分機 11,
Code Examples

Chapter 11: Pandas Package
How to read data (files: Txt, Execel, HTML) to "series" and "DataFrame" ?
CH_11_3_1_EX_1_Pandas_Reading_CSV_File.py
CH_11_3_2_EX_1_Pandas_Reading_Excel_File.py
CH_11_3_3_EX_1_Pandas_Reading_HTML_File.py

Chapter 12: Matplotlib (2D or 3D Data Visiualization)
CH_12_2_EX_1_Basic_Plot_Using_Pyplot.py
CH_12_3_EX_1_Pyplot_Figure.py
CH_12_4_EX_1_Pyplot_Subplot_Common_Usages.py

1. Supervised Learning
Chapter 2 - Training Machine Learning Algorithms for Classification
Data Sets for Data Mining
Decision Tree (J48(java version of C4.5)) with "iris.arff"
iris.data


---

How to constuct one of Decision Tree (DT)?
(Shanon Entropy)(Information Gain)
Weka with weather(YouTube)
WEKA Data Sets (weather)

How to avoid overfitting when tranning a DT?
(Pre-Pruning and Post-Pruning)


---

How to evaluate the preformance of one classifier?
How to compare the performance of several classifiers?

K-Nearesst Neightbor
Training : determining the best value of k that aci=hieves the best performance ?

Naive Bayes Classifier (Probability model)
Training : Find the conditional independent probablity of variable.
Bayesian Decision Theory
Naive Bayes Classifier (From Tom M. Mitchell)
Naive_Bayes_training_example_Tennis.htm

Linear Classifier (Vecor Space Model)
Training : Find the hyperplanes that can separate the instances of one class from the other classes
Rocchi Alogirhtm (Linear Classifier)LinearClassifier_jdwang.xls

Support Vecotr Machine (Vector Space Model)
Training: Find the support vectors that can maximumize the marge region between two classes

Chapter 3 - A Tour of Machine Learning Classifiers Using Scikit-Learn
wine.data

Chapter 4 - Building Good Training Sets – Data Preprocessing
Missing Values (NaN, Not a Number)
What do you do about "Missing Values"? (impute?)
Categorical data:(1) Nominal feature (color?) (2) ordinal feature (Size?)
Feature Scaling: Normalization vs. standardization
How to choose meaningful features? (L1 regularization & L norm2 )
Dimension Reduction:Feature selection vs. Feature extraction

Chapter 5 Dimension Reduction
Principle component analysis (PCA)(unsupervised)
Linear discriminat analysis (LDA)(Fishr's LDA) (Supervised)

Chapter 8 - Applying Machine Learning To Sentiment Analysis
Opinion mining practice: Internet Movie Database (IMDb)
(0) Download aclImdb_v1.tar.gz
(1) uncompressed "aclImdb_v1.tar.gz" into the working directory
(2) download Chapter08.7z
(3) python package: PyPrimd, NLTK
conda prompte>pip install pyprind
conda prompte>pip install nltk
Vectorization : (n-gram model)(Bag-of-words)(Vect2Word)
word stemming: Poster stemmer algorithm
Pattern Weighting:(TF)(DF)(IDF)

Chapter 9 - Web Application Embedded with Machine learning Mode
http://raschkas.pythonanywhere.com/
download Chapter09.7z
Model persistence (package : packle)
Anacoda Prompt > pip install flask
C:\Users\jdwan\Chapter09\1st_flask_app_1>python app.py * Restarting with stat * Debugger is active! * Debugger PIN: 559-490-217 * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
WTForms
Anacoda Prompt> pip install WTForms
SQLLite Manager
PythonAnywhere

Chapter 10 - Predicting Continuous Target Variables with Regression Analysis
How to predict the price of one hourse?
housing.data.txt
(506 samples in which contain 14 features)
housing.names Chapter.7z
conda prompt> pip install seaborn
Why we need the pairplot(Exploratory Data Analysis, EDA)
Perrson Product-monment correlation coefficients, Persson'r
RANdom SAmple Consensus (RANSAC)

Chapter 12 - Implementing a Multi-layer Artificial Neural Network from Scratch
MNIST dataset

Chapter 13 - Parallelizing Neural Network Training with TensorFlow

Chapter 14 - Going Deeper: The Mechanics of TensorFlow

Chapter 15 - Classifying Images with Deep Convolutional Neural Networks

Chapter 16 - Modeling Sequential Data Using Recurrent Neural Networks

2. UnSupervised Learning
Chapter 11 - Working with Unlabeled Data – Clustering Analysis
k-means
prototype-based
hierarchical-based
density-based
How many clusters ?
elbow method
hard clustering
soft clustering (or Fuzzy C-means, FCM)

3. Reinforcement Learning
4. Machine Learning Practice with Python

(1) Where were the resource derived from ?
(2) The format of thse ECG data ?
(3) What can you extract from these data?
(4) Try to losd these data using python "Panda", and to view with python "Metplotlib", and to verify (learn for classification) with python "Scikit".
(5) Find Related Potential Applications : MIT ECG signal processing, data mining, classification and clustering

Chapter 01: 建置Python 開發環境

Microsoft Docs
Microsoft Azure Portal