The open-source development model has played a pivotal role in the steady emergence of machine learning and artificial intelligence into the mainstream. Many libraries and frameworks are available to developers as open source code, and cloud computing giants like Google, Microsoft, and AWS have led the way in providing many of these projects.
One reason for the influx of open source machine learning and AI projects is that it lowers the barriers to entry for developers who can experiment and become proficient with high-quality frameworks, libraries, and applications. Most enterprises are familiar with the exciting use cases for machine learning and AI apps. In fact, 54 percent of executives are already actively investing in AI.
However, there is a huge talent gap that prevents many companies from deploying such apps in production environments. The open-source model helps, in part, to close that talent gap by encouraging developers to experiment and learn. The fact that many of the projects for AI and machine learning have been released by data-science heavyweights like Google and Amazon is important because it ensures developers become familiar with high-quality code from the outset.
With this in mind, read on to find out about seven of the most interesting open source machine learning and AI frameworks, libraries, and tools you can start working with. But first, a brief discussion on open source security.
Open Source Security
Open source components are generally quite secure, although vulnerabilities do exist. The worrying fact is not the potential for open source projects to contain vulnerabilities, but the lack of attention companies often pay to open source security. A 2015 study found that even though 78 percent of companies use open source components, 67 percent reported not monitoring open source code for security vulnerabilities.
It is reassuring that giants of the technology industry have built and released many of the open source projects for machine learning and AI, meaning they are much less likely to contain flaws. However, it’s still imperative to follow open source security best practices when using any freely available library or framework. Early detection and removal of vulnerabilities is the only way to guarantee safety.
This resource by WhiteSource goes into thorough detail on open source security and how to automatically detect vulnerabilities in libraries or frameworks.
Open Source Machine Learning/AI Tools and Frameworks
TensorFlow
TensorFlow is an open source machine learning framework created by Google. This framework is well-regarded due to its comprehensive approach; it allows you to develop neural networks using flowgraphs. TensorFlow has a Python-based interface that many developers may find they can get to grips with.
The project is well-maintained, with over 28,000 commits and over 1,300 contributors on GitHub. Other frameworks can be built on top of TensorFlow, extending its functionality. The framework is an excellent choice for sound-based use cases, such as developing speech-to-text apps, and even detecting flaws in automotive and aviation engines.
LIME
Short for Local Interpretable Model-Agnostic Explanations, LIME is a helpful Python package that intelligently explains the predictions of a given machine learning model. Being able to explain the rationale behind predictions made by neural networks is no small feat: such explanations would make you better positioned to trust or mistrust applications and their underlying predictive models.
In an interesting article by the creator of LIME, there’s an explanation given for the top three image classes predicted by Google’s Inception neural network when it analyzes a photo of a labrador playing acoustic guitar. The model mistakenly chooses an electric guitar the highest proportion of the time because the fretboard looks similar to an acoustic one. The package also works with raw text classifiers.
Accord.NET
Accord.NET is a machine learning framework written in the C Sharp (C#) programming language. This framework has multiple libraries, 40 different statistical distributions, more than 30 hypothesis tests, and more than 38 kernel functions, so you can do a lot with it.
The libraries cover tasks such as static data processing, machine learning, and pattern recognition. You can use the framework to build high-performance facial recognition apps based on deep learning.
Amazon Machine Learning
Amazon Machine Learning (AML) is an excellent choice because it’s purposefully built to help you learn the basics of creating machine learning models without delving into the dizzying complexities of how it all works under the hood .
Amazon offers tools and services for machine learning tailored to meet the needs of users and their level of expertise. The most exciting aspect is the API-drive machine learning, which lets developers easily add intelligence to any application from a selection of services providing computer vision, speech, language analysis, and chatbot functionality.
Microsoft Cognitive Toolkit
Microsoft Cognitive Toolkit is an AI framework written in C++. According to Microsoft, the toolkit trains deep learning algorithms to learn like the human brain, which is an exciting claim. This framework has been used to develop a range of models, including speech recognition, and even a model that attempts to classify whether a tumor is malignant or benign, given just two features (age and tumor size).
Caffe
Caffe is a recent addition to the list of open source machine learning frameworks. Caffe is written in C++ and has a Python-based interface in addition to a command line interface.
The main draw of this framework is its expressiveness, modularity, and speed. In terms of speed, Caffe can process over 60 million images on a daily basis with a single Nvidia K40 GPU, and it is an excellent choice for visual recognition apps that use machine learning.
Theano
Theano is one of the oldest open source machine learning libraries, having first been released in 2007. The main advantage, aside from its established reputation, is that Theano simplifies the process of defining, optimizing, and assessing the mathematical expressions that underlie machine learning models. Theano also comes with extensive code-testing capabilities.
Wrap Up
Make sure you experiment with some of these excellent open source frameworks and libraries if you want to improve your machine learning and AI expertise. And don’t forget to neglect open source security practices.