Shulou(Shulou.com)05/31 Report--

At its annual WWDC event, Apple announced the release of a sequel to its CreateML tool and its Core ML framework.

CreateML aims to simplify the creation of artificial intelligence models. In fact, because it is built in Swift, you can use a drag-and-drop programming interface like Xcode Playgrounds to train the model.

Core ML, Apple's machine learning framework, was first launched on WWDC last year. This year, the company has been working to make it leaner.

Apple claims to have improved the speed of Core ML 2 by 30 per cent using a technology called batch prediction. The quantization framework can shrink the model by up to 75%.

This is Apple's description of Core ML:

"Core ML allows you to integrate various machine learning model types into your application. In addition to supporting extensive deep learning of more than 30 layer types, it also supports standard models such as tree sets, SVM, and generalized linear models.

Based on underlying technologies such as Metal and Accelerate, Core ML can seamlessly utilize CPU and GPU to provide the highest performance and efficiency.

You can run the machine learning model on the device so that the data does not need to be analyzed by the device. "

As people become more vigilant about how to collect and use their data, they are clearly trying to iterate without leaving the device.

Google unveiled the ML toolkit at last month's input and output developer conference. Most of its functions can be run offline, but they are more limited than when connecting to the Google cloud. For example, a version of API on a device can detect dogs in photos-but when connected to the Internet-it can identify specific breeds.

In the past, developers of Memrise, a language learning app, took 24 hours to train a model containing 20000 images, Apple said. ML and Core ML 2 were shortened to 48 minutes on MacBook Pro and 18 minutes on iMac Pro. In addition, the size of the model has been reduced from 90MB to 3MB.

For developers who like Core ML but use TensorFlow, Google released a tool in December 2017 that converts AI models into compatible file types. You can find it on Microsoft's GitHub.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.