scikit-learn / keras basic implementation...
Post on 23-Sep-2020
15 Views
Preview:
TRANSCRIPT
Scikit-learn / Keras Basic Implementation Tutorial
2019.03.20Jacky, Chun-Yen Yeh
!1
Goal
• Introduce Scikit-learn, Keras Python library(framework).
• Go through the workflow of the classification problem.
• Lead you to step-by-step implement classification problem with Scikit-learn / Keras.
!2
Overview
!3
Artificial IntelligenceMachine Learning
Deep LearningDecision Tree
Nearest NegihborsLogistic Regression CNN
RNNTf-idf
Data Engineering . . .
.
.
.
Overview
!4
Artificial IntelligenceMachine Learning
Deep Learningdecision tree
Nearest NegihborsLogistic Regression CNN
RNNTf-idf
Data Engineering . . .
.
.
.
Scikit-learn
Keras
Scikit-learn
!5
What’s Scikit-learn?
!6
• A free software machine learning library for the Python language.
• Simple and efficient tools for data mining and data analysis.
• Derived from SciPy, which is a Python-based ecosystem of open-source software for mathematics, science, and engineering. (e.g. Numpy, pandas, jupyter notebook)
Why’s Scikit-learn• Commitment to documentation and usability.
• Covers most machine-learning tasks: classification, regression, clustering, dimension reduction, data preprocessing, etc…
!7
!8
Supervised Classification
!9
Species Features
Iris Dataset
Split the Dataset
!10
Training Data Testing Data
ALL Data
Typically 75% : 25% = 3 : 1
!11
Training Data
Supervised Workflow
Model
Prediction
Evaluation
Training phase
Inference phase
Training LabelTraining Labels
Testing Data
Testing Labels
Model Construction
!12
Training Data
Supervised Workflow
Model
Prediction
Evaluation
Training phase
Inference phase
Training LabelTraining Labels
Testing Data
Testing Labels
Model Construction
classifier.fit(X_train, y_train)
classifier = LogisticRegression()
classifier.predict(X_test)
classifier.score(X_test, y_test)
Exercise time
!13
Please enter this file and click your link. https://reurl.cc/nN28n
Keras
!14
What’s Keras?
!15
Keras is a high-level deep learning API, written in Python and capable of running on top of TensorFlow, CNTK, or Theano.
Why’s Keras?
• Allows for easy and fast prototyping (through user friendliness, modularity, and extensibility).
• Supports both convolutional networks and recurrent networks, as well as combinations of the two.
• Runs seamlessly on CPU and GPU.
• Complete documentation (keras.io)
!16
Three Steps of Deep Learning
!17
Step1 Define a set of
function
Step2 Goodness of
function
Step3 Pick the best
function
!18Output
Like lego :)
!19
Prediction Real Label
loss (difference)
!20
parameters
choose optimizer (SGD, RMSProp, …)
!21
Training Data
Supervised Workflow
Model
Prediction
Evaluation
Training phase
Inference phase
Training LabelTraining Labels
Testing Data
Testing Labels
Model Construction
classifier.predict_classes(X_test)
classifier.score(X_test, y_test)
Exercise time
!22
Please enter this file and click your link. https://reurl.cc/nN28n
!23
top related