Book Image

Machine Learning with Swift

By : Jojo Moolayil, Alexander Sosnovshchenko, Oleksandr Baiev
Book Image

Machine Learning with Swift

By: Jojo Moolayil, Alexander Sosnovshchenko, Oleksandr Baiev

Overview of this book

Machine learning as a field promises to bring increased intelligence to the software by helping us learn and analyse information efficiently and discover certain patterns that humans cannot. This book will be your guide as you embark on an exciting journey in machine learning using the popular Swift language. We’ll start with machine learning basics in the first part of the book to develop a lasting intuition about fundamental machine learning concepts. We explore various supervised and unsupervised statistical learning techniques and how to implement them in Swift, while the third section walks you through deep learning techniques with the help of typical real-world cases. In the last section, we will dive into some hard core topics such as model compression, GPU acceleration and provide some recommendations to avoid common mistakes during machine learning application development. By the end of the book, you'll be able to develop intelligent applications written in Swift that can learn for themselves.
Table of Contents (18 chapters)
Title Page
Packt Upsell

Compact CNN architectures

During the inference, the whole neural network should be loaded into the memory, so as mobile developers we are especially interested in the small architectures, which consume as little memory as possible. Small neural networks also allow to reduce the bandwidth consumption when downloaded from the network.

Several architectures designed to reduce the size of convolutional neural networks have been proposed recently. We will discuss in brief several most known of them.


The architecture was proposed by Iandola et al. in 2017 for use in autonomous cars. As the baseline, researchers took the AlexNet architecture. This network takes 240 MB of memory, which is pretty much the equivalent of mobile devices. SqueezeNet has 50x fewer parameters, and achieves the same level of accuracy on the ImageNet dataset. Using additional compression, its size can be reduced to about 0.5 MB.

SqueezeNet is built from the fire modules. The objective was to create a neural network...