やはり、御三家資格(Cisco, Oracle, MS On-Premise)はどんどん人気がなくなり、Cloud系が台頭。特にAWS!
本 - 人工知能の未来 2019 - 2023
高すぎて買わないけど、目次を眺めるだけでも、AI全体を鳥瞰するのにいいかも~。
洋書 - Modular Programming with Python
Modular Programming with Python
Modular Programming with Python
- 作者: Erik Westra
- 出版社/メーカー: Packt Publishing
- 発売日: 2016/05/26
- メディア: ペーパーバック
- この商品を含むブログを見る
Reference
感想
実際pipからダウンロードできるまでの過程がわかりやすく書いてある。
特に8章はお勧め。
Table of Contents
1. Introducing Modular Programming
2. Writing Your First Modular Program
3. Using Modules and Packages
4. Using Modules for Real-World Programming
5. Working with Module Patterns
6. Creating Reusable Modules
7. Advanced Module Techniques
8. Testing and Deploying Modules
8. Testing and Deploying Modules
- Testing modules and packages
- Preparing a module or package for publication
- Uploading your work to GitHub
- Submitting to the Python Package Index
- https://pypi.python.org/pypiから!
- LICENSE.txtの例 : MIT license
- Using pip to download and install modules and packages
9. Modular Programming as a Foundation for Good Programming Technique
Python - Matplotlib参考情報
Link
Official
- matplotlib.org
- matplotlib.org - User's Guide
- matplotlib.org - Gallery
- 各グラフをクリックすると、コードが見れる!
Others
参考書籍
- 作者: Srinivasa Rao Poladi
- 出版社/メーカー: Packt Publishing
- 発売日: 2018/10/23
- メディア: Kindle版
- この商品を含むブログを見る
Machine Learningの有名なサンプルデータ
Iris dataset
Dataset概要
- For supervised learning, classification (three-classes)
- target (classes) : setosa(0), versicolor(1), virginica(2)
- feature : sepal length (cm), sepal width (cm), petal length (cm), petal width (cm)
Technical Info
- import : from sklearn.datasets import load_iris
- dict_keys(['data', 'target', 'target_names', 'DESCR', 'feature_names', 'filename'])
- iris_dataset['data'].shape : (150, 4)
- iris_dataset['target'].shape : (150,)
取り上げている書籍
- [Introduction to Machine Learning with Python]
使われ方例
k-nearest neighbors classification algorithm
X_train, X_test, y_train, y_test = train_test_split( iris_dataset['data'], iris_dataset['target'], random_state=0) knn = KNeighborsClassifier(n_neighbors=1) knn.fit(X_train, y_train) print("Test set score: {:.2f}".format(knn.score(X_test, y_test)))
Wisconsin Breast Cancer dataset
Dataset概要
- Cancer data
Technical Info
- cancer.keys(): dict_keys(['data', 'target', 'target_names', 'DESCR', 'feature_names', 'filename'])
Boston Housing dataset
Dataset概要
Technical Info
洋書 - Introduction to Machine Learning with Python
Introduction to Machine Learning with Python
Introduction to Machine Learning with Python: A Guide for Data Scientists
- 作者: Andreas C. Mueller,Sarah Guido
- 出版社/メーカー: O'Reilly Media
- 発売日: 2016/10/21
- メディア: ペーパーバック
- この商品を含むブログを見る
Table of Contents
1. Introduction
1.1. Why Machine Learning?
1.2. Why Python?
1.3. scikit-learn
1.4. Essential Libraries and Tools
Jupyter Notebook, NumPy, SciPy, matplotlib, pandas, mglearn
1.5. Python 2 Versus Python 3
1.6. Versions Used in this Book
1.7. A First Application: Classifying Iris Species
Iris datasetを使ったk-NN法によるIris品種予測
2. Supervised Learning
2.1. Classification and Regression
2.2. Generalization, Overfitting, and Underfitting
2.2.1. Relation of Model Complexity to Dataset Size
2.3. Supervised Machine Learning Algorithms
2.3.1. Some Sample Datasets
2.3.2. k-Nearest Neighbors (k近傍法)
2.3.3. Linear Models (線形モデル)
2.3.4. Naive Bayes Classifiers (単純ベイズ分類器)
2.3.5. Decision Trees (決定木)
2.3.6. Ensembles of Decision Trees (アンサンブル)
2.3.7. Kernelized Support Vector Machines
2.3.8. Neural Networks (Deep Learning)
2.4. Uncertainty Estimates from Classifiers
2.4.1. The Decision Function
2.4.2. Predicting Probabilities
2.4.3. Uncertainty in Multiclass Classification
3. Unsupervised Learning and Preprocessing
3.1. Types of Unsupervised Learning
3.2. Challenges in Unsupervised Learning
3.3. Preprocessing and Scaling
3.3.1. Different Kinds of Preprocessing
3.3.2. Applying Data Transformations
3.3.3. Scaling Training and Test Data the Same Way
3.3.4. The Effect of Preprocessing on Supervised Learning
3.4. Dimensionality Reduction, Feature Extraction, and Manifold Learning
3.4.1. Principal Component Analysis (PCA) (主成分分析)
3.4.2. Non-Negative Matrix Factorization (NMF) (非負値行列因子分解)
3.4.3. Manifold Learning with t-SNE
3.5. Clustering
3.5.1. k-Means Clustering (k平均法)
3.5.2. Agglomerative Clustering (凝集型クラスタリング)
3.5.3. DBSCAN
3.5.4. Comparing and Evaluating Clustering Algorithms
3.5.5. Summary of Clustering Methods
4. Representing Data and Engineering Features
4.1. Categorical Variables
4.1.1. One-Hot-Encoding (Dummy Variables) (One-Hot表現)
4.1.2. Numbers Can Encode Categoricals
4.2. OneHotEncoder and ColumnTransformer: Categorical Variables with scikit-learn
4.3. Convenient ColumnTransformer creation with make_columntransformer
4.4. Binning, Discretization, Linear Models, and Trees
4.5. Interactions and Polynomials
4.6. Univariate Nonlinear Transformations (単変量非線形変換)
4.7. Automatic Feature Selection
4.7.1. Univariate Statistics
4.7.2. Model-Based Feature Selection
4.7.3. Iterative Feature Selection
4.8. Utilizing Expert Knowledge
5. Model Evaluation and Improvement
5.1. Cross-Validation
5.1.1. Cross-Validation in scikit-learn
5.1.2. Benefits of Cross-Validation
5.1.3. Stratified k-Fold Cross-Validation and Other Strategies
5.2. Grid Search
5.2.1. Simple Grid Search
5.2.2. The Danger of Overfitting the Parameters and the Validation Set
5.2.3. Grid Search with Cross-Validation
5.3. Evaluation Metrics and Scoring
5.3.1. Keep the End Goal in Mind
5.3.2. Metrics for Binary Classification
5.3.3. Metrics for Multiclass Classification
5.3.4. Regression Metrics
5.3.5. Using Evaluation Metrics in Model Selection
6. Algorithm Chains and Pipelines
6.1. Parameter Selection with Preprocessing
6.2. Building Pipelines
6.3. Using Pipelines in Grid Searches
6.4. The General Pipeline Interface
6.4.1. Convenient Pipeline Creation with make_pipeline
6.4.2. Accessing Step Attributes
6.4.3. Accessing Attributes in a Pipeline inside GridSearchCV
6.5. Grid-Searching Preprocessing Steps and Model Parameters
6.6. Grid-Searching Which Model To Use
6.6.1. Avoiding Redundant Computation
7. Working with Text Data
この章、相当難易度が高い・・・
7.1. Types of Data Represented as Strings
7.2. Example Application: Sentiment Analysis of Movie Reviews
7.3. Representing Text Data as a Bag of Words
7.4. Stopwords
7.5. Rescaling the Data with tf–idf
7.6. Investigating Model Coefficients
7.7. Bag-of-Words with More Than One Word (n-Grams)
7.8. Advanced Tokenization, Stemming, and Lemmatization
7.9. Topic Modeling and Document Clustering
Latent Dirichlet Allocation
8. Wrapping Up
8.1. Approaching a Machine Learning Problem
8.1.1. Humans in the Loop
8.2. From Prototype to Production
8.3. Testing Production Systems
8.4. Building Your Own Estimator
8.5. Where to Go from Here
8.5.1. Theory
8.5.2. Other Machine Learning Frameworks and Packages
8.5.3. Ranking, Recommender Systems, and Other Kinds of Learning
8.5.4. Probabilistic Modeling, Inference, and Probabilistic Programming
8.5.5. Neural Networks
8.5.6. Scaling to Larger Datasets
8.5.7. Honing Your Skills
8.6. Conclusion
洋書 - Hands-On Data Analysis with NumPy and pandas
Hands-On Data Analysis with NumPy and pandas
- 作者: Curtis Miller
- 出版社/メーカー: Packt Publishing
- 発売日: 2018/06/29
- メディア: Kindle版
- この商品を含むブログを見る
感想
紙の本だと168ページぐらい。 何よりわかりやすい(内容も深くない)。
いきなりMachine Learningのアルゴリズムに挑戦して撃沈(!)した場合、基本に戻ってNumPy、pandasの基礎を学ぶのにいい。
Table of Contents
1. SETTING UP A PYTHON DATA ANALYSIS ENVIRONMENT
2. DIVING INTO NUMPY
- NumPy arrays
- Special numeric values
- Creating NumPy arrays
- Creating ndarray
3. OPERATIONS ON NUMPY ARRAYS
- Operations on NumPy Arrays
- Selecting elements explicitly
- Slicing arrays with colons
- Advanced indexing
- Expanding arrays
- Arithmetic and linear algebra with arrays
- Arithmetic with two equal-shaped arrays
- Broadcasting
- Linear algebra
- Employing array methods and functions
- Array methods
- Vectorization with ufuncs
- Custom ufuncs
4. PANDAS ARE FUN! WHAT IS PANDAS?
- What does pandas do?
- Exploring series and DataFrame objects
- Creating series
- Creating DataFrames
- Adding data
- Saving DataFrames
- Subsetting your data
- Subsetting a series
- Indexing methods
- Slicing a DataFrame
5. ARITHMETIC, FUNCTION APPLICATION, AND MAPPING WITH PANDAS
- Arithmetic, Function Application, and Mapping with pandas
- Arithmetic
- Arithmetic with DataFrames
- Vectorization with DataFrames
- DataFrame function application
- Handling missing data in a pandas DataFrame
- Deleting missing information
- Filling missing information
6. MANAGING, INDEXING, AND PLOTTING
- Managing, Indexing, and Plotting
- Index sorting
- Sorting by values
- Hierarchical indexing
- Slicing a series with a hierarchical index
- Plotting with pandas
- Plotting methods
洋書 - TensorFlow for Deep Learning
TensorFlow for Deep Learning
TensorFlow for Deep Learning: From Linear Regression to Reinforcement Learning
- 作者: Bharath Ramsundar,Reza Bosagh Zadeh
- 出版社/メーカー: O'Reilly Media
- 発売日: 2018/03/23
- メディア: ペーパーバック
- この商品を含むブログを見る
Table of Contents
1. Introduction to Deep Learning
2. Introduction to TensorFlow Primitives
3. Linear and Logistic Regression with TensorFlow
4. Fully Connected Deep Networks
5. Hyperparameter Optimization
6. Convolutional Neural Networks
7. Recurrent Neural Networks
8. Reinforcement Learning
9. Training Large Deep Networks
10. The Future of Deep Learning
AI Boomの現状 - 9つのチャート (2018/12)
Link
主要ポイント
- AI is being commercialized at a dizzying pace.
- The focal points are China and the US, but also Europe.
- There are still far more men in the field than women.
- The state of the art is improving fast.
- Artificial intelligence is a political issue.
Python - Deep Learning / TensorFlow参考情報
Deep Learning参考ビデオ・書籍
Lynda.com
- Building and Deploying Deep Learning Applications with TensorFlow (by Adam Geitgey)
- Neural Networks and Convolutional Neural Networks Essential Training (by Jonathan Fernandes)
Safari Online
- Fundamentals of Deep Learning (by Nikhil Buduma, Publisher: O'Reilly Media, Inc.)
- 相当理論的な本。参考書として使ったほうがいいかも
- TensorFlow for Deep Learning (by Reza Bosagh Zadeh, Bharath Ramsundar, Publisher: O'Reilly Media, Inc.)
- 実践的な本。サンプルを試しながら理解していく本。
その他リンク
Amazonからの購入
Fundamentals of Deep Learning: Designing Next-Generation Machine Intelligence Algorithms
- 作者: Nikhil Buduma,Nicholas Locascio
- 出版社/メーカー: O'Reilly Media
- 発売日: 2017/06/29
- メディア: ペーパーバック
- この商品を含むブログ (1件) を見る
実践 Deep Learning ―PythonとTensorFlowで学ぶ次世代の機械学習アルゴリズム (オライリー・ジャパン)
- 作者: Nikhil Buduma,太田満久,藤原秀平,牧野聡
- 出版社/メーカー: オライリージャパン
- 発売日: 2018/04/26
- メディア: 単行本(ソフトカバー)
- この商品を含むブログ (2件) を見る
TensorFlow for Deep Learning: From Linear Regression to Reinforcement Learning
- 作者: Bharath Ramsundar,Reza Bosagh Zadeh
- 出版社/メーカー: O'Reilly Media
- 発売日: 2018/03/23
- メディア: ペーパーバック
- この商品を含むブログを見る