Apple Machine Learning
- 1. Apple Machine Learning Denise Nepraunig
- 2. Do Machine Learning they said. It will be fun they said.
- 3. Statistics Algebra Matrices Vectors Eigenvectors Multivariable Calculus Gradient Descent Eigenvalues Python Pandas scikit-learn Neural Networks Tensorflow Jupyter Notebooks Support Vector Machines Naïve Bayes Decision Tree Classifier Boosted Trees Regression R Matlab Logistic Regression Random Forest Classifier RNN LSTM CNN NLP
- 6. Create ML Released 2018 Part of Xcode 10 Machine Learning without additional tools, simple and easy to use.
- 8. Regression Predict a number, e.g. house prices Which Regression should you use? • Logistic Regression • Boosted Trees Regression • Random Forest Regression • Decision Tree Regression • … ??? weight, price
- 9. Let the framework choose the best Regression for you! The framework analyzes the data and chooses the best model automatically. Nevertheless it is still possible to use a MLLogisticRegression, MLRandomForestRegression, MLBoostedTreesRegression ... Model can be saved with on line of code and used in Xcode later.
- 10. Text classification Predict a label, e.g. categories or spam food, rating tasty, disgusting
- 11. Create ML – Image Classifier Training Data Used 12 images for each category
- 12. Create ML – Image Classifier Xcode 10 Mac OS Playground drag & drop
- 13. drag & drop
- 16. IMG_2449.jpeg
- 18. Turi Create Python Framework Mission statement: You don't have to be a machine learning expert to add recommendations, object detection, image classification, image similarity or activity classification to your app.
- 20. Turi Create Very easy to train machine learning models as a developer. No math involved, no tuning of parameters for the model needed, high abstraction level, Python coding is readable even if you don’t know Python, model can be exported to be used in Xcode. SFrame is a tabular data structure which can also contain images
- 22. Image Classifier An Image Classifier is not helpful for analyzing such a picture.
- 23. Object Detection Object detection is the task of simultaneously classifying (what) and localizing (where) object instances in an image.
- 24. Object Detection - data preparation
- 25. Images and annotations have been loaded and trained with an object detection model (see previous slides). Afterwards the model was loaded and used to inspect the shown pictures. The model can also be exported to be used in Xcode.
- 27. Thank you!