Apple released the Core ML framework in iOS11, making it simple to integrate machine learning models into your app. In this workshop, you will learn the lifecycle of training, deploying, and evaluating a machine learning model for making on device predictions.
Compute power, large datasets, time, and deep expertise are only perceived barriers, and there are plenty of options for overcoming them! Using techniques such as transfer learning and taking inspiration from the plentiful resources from industry and academia, you can get started doing this now. Note: We will move beyond the common cat/dog classifier.
As a reminder, predictions that have traditionally been made server side can now be made locally - offering unique advantages to the user, such as more data privacy and lower latency. Hardware improvements targeted for performing machine learning tasks (iPhone X’s A11 chip), industry standard models optimized for mobile (MobileNets), and many more recent developments signal that machine learning on mobile is part of the new path forward.
Target Audience:
• Experience building iOS apps in Swift
No previous machine learning experience necessary.
What we will cover:
• Brief overview of ML terminology, tasks, and common algorithms
• Typical end-to-end machine learning pipeline and where your iOS app fits into it
• Identifying areas of your app that could be helped with machine learning & clearly defining the problem(s) to be modeled
• Acquiring a suitable dataset
• Training a machine learning model using Tensor Flow. We'll also discuss other options such as Keras & Apple's TuriCreate open source training framework.
• Deploying the trained model to Core ML
• Integrating the Core ML model into your app to make predictions on device.
• Using Core ML in combination with Vision when performing computer vision tasks
• Evaluating the performance of your machine learning model
Meghan is a software developer at Novoda in Berlin, focusing on iOS development. She is the co-creator and instructor of a Udacity course on Core ML. In the past, she has worked at SoundCloud, The Knot, and the MIT Media Lab. She enjoys the cross section of software and scientific research and hopes to do more in that space in the future. In 2017, she had the opportunity contribute to a white paper for the vision of high energy physics (HEP) software in the 2020s, collaborating with physicists from CERN (& other labs) https://arxiv.org/abs/1712.07959.
When not coding, she can usually be found reading at a coffee shop or cycling. In her former life, she lived in the US and studied Mathematics & CS at MIT.
