WWDC 2017
Table of Contents
=================
* [Using Metal 2 for Compute \- Thursday](#using-metal-2-for-compute---thursday)
* [MPS (Metal Performance Shader)](#mps-metal-performance-shader)
* [Image Processing](#image-processing)
* [Linear Algebra](#linear-algebra)
* [Accelerating ML Using MPS](#accelerating-ml-using-mps)
* [Training & Inference](#training--inference)
* [Convolutional Neural Networks](#convolutional-neural-networks)
* [Neural Network Graph API](#neural-network-graph-api)
* [Summary](#summary)
# Using Metal 2 for Compute - Thursday
Session video and resources: https://developer.apple.com/videos/play/wwdc2017/608/
## MPS (Metal Performance Shader)
- MPS is now supported on macOS as well
### Image Processing
### Linear Algebra
## Accelerating ML Using MPS
### Training & Inference
- Training to classify images -> Inference via CNN (Convolutional Neural Network)
- First part is one time, highly time consuming training phase
- Second part could be accelerated by using MPS
## Convolutional Neural Networks
## Neural Network Graph API
- Describe neural networks using graph API
- Initialize once, reuse
- Execute graph on GPU with single call
## Summary
- GPU accelerated primitives
- Extended support for image processing
- Addes support for linear algebra
- Optimized for iOS and macOS
- New neural network graph API