But nowadays every time I see ML on HN, it happens to mean ML = Machine Learning instead of ML = programming language.
Another use case is privacy: ship a pretrained facenet style model and let it learn to map your address book to faces in your photos without your data leaving the device.
I'd assume Apple does not allow this.
However it could be a killer feature to have easy api's for doing personalization all on device.
This is major, if they have managed to achieve it reasonably. But before opening a Sekt, I want to see some benchmarks. :)
How they actually compare?
Also, some converted Core ML Models ready to use here: developer.apple.com/machine-learning
A bit disappointed that the model conversion tool only supports an older version of Keras as well (1.2.2). Keras 2.0 is pretty new but I hope they update the conversion tools for it quickly ...
I wonder if the conversion tool will be open source ... seems like they'd want to support the widest net of external models since they don't yet have a way to produce .coreml models directly. Or maybe the intent is to augment Keras/caffe/etc to support saving .coreml directly?
Right now, coremltools is only available for Python 2.7 (https://pypi.python.org/pypi/coremltools), which is annoying as the entire code base I've worked on for months at my current firm is in Python 3.6. Hopefully this is updated soon for Python 3 support.
So, same answer.
And why mention libsvm, are we in the 90s.