CoreML - Machine Learning in iOS11 and Swift 4

1 minute read

For my most recent lab session, I spent the afternoon exploring the new machine learning API available. The ease at which apple is allowing you to add local (to the device) model processing is quite an exception.

CoreML allows you to load pre-trained models generated from a number of machine learning engines. These must be uploaded as a .mlmodel file. This then gives you the ability to make predictions. Basically CoreML can handle Classification and regression tasks.

CoreMLTasks.png

There are limitations from the platform but I imagine these will be ironed out in future releases of the platform. The model that is used is a static one, this means that you cannot train the model on the device with data that is gathered by the app. Also, the model is not encrypted so if you have a proprietary data model or sensitive data stored then you may want to seek an alternative.

As it goes though, I think that is not bad and the results are pretty amazing.

This new platform definitely warrants more investigation.