While many factors have contributed to this increase in machine learning, one reason is that it’s becoming easier for developers to apply it, thanks to open source frameworks.
If you’re not familiar with this technology, and feel confused about some of the terms used, such as “framework” and “library,” here are the definitions:
Framework. A vague term, to be sure; even those who regularly use it can’t agree on its exact definition. However, in most cases, “framework” refers to a bunch of programs, libraries and languages you have built to use in application development. Think of a framework as a base for getting started.
Library. A collection of objects or methods that your application uses. It’s a file with re-usable code that can be shared by many applications, so you don’t have to write the same code repeatedly. Instead, you link to the library.
As one online user put it: “The key difference between a library and a framework is ‘inversion of control.’ When you call a method from a library, you are in control. But with a framework, the control is inverted: The framework calls you.”
Still confused? Check out this helpful YouTube video about the difference between a framework and a library.
If you’re diving into machine learning in a big way, you’re probably seeking resources to help guide you. There are many frameworks available, but here are some of our favourites to help you get started.
The machine learning resources you’ll use
TensorFlow. TensorFlow was developed by the Google Brain Team to handle perceptual and language understanding tasks. It can also conduct research on machine learning and deep neural networks. TensorFlow has a Python-based interface. It’s used in many of Google’s products, handling speech recognition, Gmail, photos and search.
What’s useful about this framework is that it can perform elaborate mathematical computations and see data flow graphs. TensorFlow is flexible, meaning users can write their own libraries on top of it. It’s also portable, able to run in the cloud and on mobile computing platforms as well as with CPUs or GPUs.
Amazon Machine Learning. Amazon Machine Learning (AML) is built for developers, with many tools and wizards to help you create machine learning models without having to learn all the complexities of how machine learning works. With AML, you can generate predictions and use data from Amazon Redshift, the data warehouse Platform as a Service.
Shogun. Shogun has many state-of-the-art algorithms, making it a handy tool. It is written in C++ and provides data structures for machine learning problems. It can run on Windows, Linux and MacOS. Further, Shogun is helpful because it supports bindings to other machine learning libraries. The list is extensive, but they include: SVMLight, LibSVM, libqp, SLEP, LibLinear, VowpalWabbit and Tapkee.
Accord.NET. Accord.NET, a NET machine learning framework, has multiple libraries to handle everything from pattern recognition, image and signal processing to linear algebra, statistical data processing and more. Accord is useful because it has so much to offer, including 40 different statistical distributions, more than 30 hypothesis tests, and more than 38 kernel functions.
Apache Signa, Apache Spark MLlib and Apache Mahout. Apache Signa, Apache Spark MLlib, and Apache Mahout are three frameworks with a lot to offer. Apache Signa is mostly used in natural language processing and image recognition; it can run over a wide range of hardware.
Mahout provides Java libraries and Java collections for various kinds of mathematical operations. Spark MLlib was created with the goal of making machine learning easy. It brings together many learning algorithms and utilities, including classification, clustering, dimensionality reduction and more.
The data sets you’ll need
Once you get going, you’ll also need some data. If you’re just learning and need to practice, here are some useful data sets to try, all available on GitHub:
- Medical data
- Mixture of data sets
- National Park Service
- Google Trends
- General Services Administration (GSA)