Maschinelles Lernen mit MATLAB

Transcription

1 Maschinelles Lernen mit MATLAB Jérémy Huard Applikationsingenieur The MathWorks GmbH 2015 The MathWorks, Inc. 1

2 Machine Learning is Everywhere Image Recognition Speech Recognition Stock Prediction Medical Diagnosis Data Analytics Robotics and more [TBD] 2

3 Machine Learning Machine learning uses data and produces a program to perform a task Task: Human Activity Detection Standard Approach Machine Learning Approach Computer Program Machine Learning Hand Written Program If X_acc > 0.5 then SITTING If Y_acc < 4 and Z_acc > 5 then STANDING Formula or Equation Y activity = β 1 X acc + β 2 Y acc + β 3 Z acc + model:inputs Outputs model = < Machine@Learning@Algorithm >(sensor_data, activity) 3

4 Different Types of Learning Machine Learning Supervised Learning Unsupervised Learning Discover a good internal representation Learn a low dimensional representation Classification Output is a choice between classes (True, False) (Red, Blue, Green) Regression Output is a real number (temperature, stock prices). 4

5 Essentially, all models are wrong, but some are useful George Box 5

6 General Challenges in Machine Learning Hard to get started Steps Access, explore and analyze data Preprocess data Train models Assess model performance Iterate Challenge Data diversity Numeric, Images, Signals, Text not always tabular Lack of domain tools Filtering and feature extraction Feature selection and transformation Time consuming Train several models to find the best Avoid pitfalls Over Fitting Speed-Accuracy-Complexity tradeoffs 6

7 Supervised Learning Workflow Train: Iterate till you find the best model LOAD DATA PREPROCESS DATA SUPERVISED LEARNING MODEL FILTERS PCA CLASSIFICATION SUMMARY STATISTICS CLUSTER ANALYSIS REGRESSION Predict: Integrate trained models into applications NEW DATA PREDICTION 7

8 Statistics and Machine Learning What s New Classification Learner New app to train models and classify data using supervised machine learning Features Import and interactively explore data Choose kfold or holdout validation Train SVM, knn, bagged trees and other algorithms Assess results using classification accuracy, ROC curves and Confusion Matrices Export models to the MATLAB or generate MATLAB code 8

9 Train a Model with the Classification Learner App 9

10 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 10

11 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 11

12 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 12

13 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 13

14 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 14

15 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 4. Model comparison and assessment 15

16 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 4. Model comparison and assessment 5. Share model 16

17 Train a Model with the Classification Learner App 1. Data import and Crossvalidation setup 2. Data exploration and feature selection 3. Train multiple models 4. Model comparison and assessment 5. Share model or automate process 17

18 Statistics and Machine Learning What s New? New: Classification Learner app Multiclass SVM Statistical tests for comparing classifiers Kmediods Clustering (robust to outliers) C Code Generation for PCA Requires MATLAB Coder Enhancements: Speedup of the kmeans and gmdistribution using the kmeans++ Performance enhancements for decision trees and performance curves 18

19 Key Takeaways Machine Learning with MATLAB: For complex tasks with no equation or formula Interactive App-driven workflow Flexible architecture for customized workflow 19

20 Additional Resources Documentation mathworks.com/machine-learning Training 20