Today, everywhere we look, Machine Learning is around us in some form or the other. This subset of Artificial Intelligence has found diverse applications across all parallels of the industry, and rightly so. Even though Machine Learning is an emerging field, it has opened up a slew of possibilities to explore.
Now, the question is, which programming language to use for Machine Learning projects?
Python and C++ are two of the most popular programming languages. Both of these languages boast of an active community, dedicated tool support, an extensive ecosystem of libraries, and commendable runtime performance. However, the focus of today’s post is going to be Machine Learning in C++.
Why C++ for Machine Learning?
It is a well-established fact that Machine Learning requires heavy-duty CPU performance, and this is precisely what C++ guarantees. When it comes to speed and performance, C++ leaves behind Python, Java, and even C#. Another major advantage of using C++ for Machine Learning is that it has pointer support, a feature not available in many of the popular programming languages.
For the successful implementation of Machine Learning in C++, the foremost thing to do is to acquaint yourself with C++ libraries. Thankfully, C++ has some great libraries for Machine Learning, including Shark, MLPack, and GRT (Gesture Recognition Toolkit).
Now, let’s dive into the discussion of Machine Learning libraries in C++.
Machine Learing Libraries in C++
Shark is an open-source, modular library in C++. It is the perfect library for Machine Learning since it has extensive support for supervised learning algorithms like linear regression, k-means, neural networks, and clustering, to name a few.
Shark also includes numerous methods for linear and nonlinear optimization, kernel-based learning algorithms, numerical optimization, and a host of other ML techniques. It is the ideal tool for both research and building real-world applications. Shark has excellent documentation and is compatible with Linux, Windows, and macOS.
How to install Shark?
To install Shark, you have to get the source packages from the official downloads page. After this, you must build the library by writing the following code:
You must know that Shark has two dependencies – Boost and CMake. While on Linux ad Windows, usually ATLAS is used, on macOS, Accelerate is the default linear algebra library. In macOS, you can use MacPorts to obtain the necessary packages, like so:
sudo port install boost cmake
However, under Ubuntu, you have to install the required packages by using the following statement:
sudo apt-get install cmake cmake-curses-gui libatlas-base-dev libboost-all-dev
Here are the steps for installing Shark:
- First, download the source packages from the downloads page and unpack them.
- Launch the CMake GUI
- Select “Where is the source code” to set the path to the unpacked Shark location.
- Select “Where to build the directory” to set the path where you want to store the Visual Studio project files.
- Choose the “Add Entry” option. Now, add an Entry BOOST_ROOT of type PATH and set it to your boost install directory.
- Again, add an Entry BOOST_LIBRARYDIR of type PATH and set it to your boost library directory.
- Finally, choose the apt Visual Studio compiler and double-click on the “Configure” option, followed by the “Generate” option.
mlpack is a C++ library that is designed explicitly for performance. It promises to offer fast and extensible implementations of pioneering ML algorithms. The unique aspect of this C++ library is that it provides the ML algorithms as simple command-line programs, Python bindings, Julia bindings, and C++ classes, all of which you can integrate into larger-scale ML solutions.
How to install mlpack?
The installation process of MLPack varies from platform to platform.
For Python, you can get the source package through pip or conda, like so:
pip install mlpack
conda install -c conda-forge mlpack
You can refer to the mlpack in Python quickstart guide for more details.
For Julia, you can get the sources via Pkg, as follows:
For Ubuntu, Debian, Fedora, and Red Hat, you can install mlpack using a package manager. The mlpack command-line quickstart guide is a good place to start. You can also build it from source following the Linux build tutorial.
For Windows, you can download prebuilt binaries – Windows 64 bit – MSI Installer and Windows 64 bit – ZIP. You can also install it using a package manager like vcpkg, or build from source following the Windows build tutorial.
Coming to macOS, you can install the library via homebrew, like so:
brew install mlpack
3. GRT (Gesture Recognition Toolkit)
GRT or Gesture Recognition Toolkit is an open-source, cross-platform C++ library. It is specially designed for real-time gesture recognition. It includes a comprehensive C++ API that is further solidified by a neat and easy-to-use GUI (Graphical User Interface).
GRT is not only beginner-friendly, but it is also extremely easy to integrate into existing C++ projects. It is compatible with any sensor/data input, and you can train it with your unique gestures. Furthermore, GRT can adapt to your custom processing or feature extraction algorithms as and when needed.
How to install GRT?
The first thing you must do is to download the GRT package. After this, you must locate the GRT folder in the main gesture-recognition-toolkit folder and add the GRT folder (including all the subfolders) to the desired project.
You can start using the GRT by adding the complete code stored in the GRT folder to your C++ project. In case you use IDEs like VisualStudio or XCode, you can add the GRT folder files to your project following this path – “File -> Add Files to project.” You can also drag the GRT folder (from Finder or Windows Explorer) into the IDE to add all the files from the GRT folder to your project.
Once you’ve added the code contained in the GRT folder to your project, you can use all the GRT functions/classes. All you have to do is add the following two lines of code to the top of the header file in the project where you want to use the GRT code:
int main (int argc, const char * argv)
//The main code for your project…
In this code, the first line adds the main GRT header file (GRT.h) to the project. The GRT.h file contains all of the GRT module header files, and hence, you don’t have to enter any other GRT header files manually. However, the second line declares that the GRT namespace is being used. This eliminates the need to write GRT:: WhatEverClass each time you want to use a GRT class – you can write WhatEverClass and be done with it.
However, remember that you have to specify the physical path where you stored the GRT folder on your hard drive, based on the IDE you use.
Also Read: Career in Machine Learning
These three C++libraries are perfect to handle almost all your ML needs. The key to mastering Machine Learning in C++ is first to learn these libraries, understand their specialties and functions, and then implement them to specific ML requirements.
If you’re interested to learn more about machine learning, check out IIIT-B & upGrad’s PG Diploma in Machine Learning & AI which is designed for working professionals and offers 450+ hours of rigorous training, 30+ case studies & assignments, IIIT-B Alumni status, 5+ practical hands-on capstone projects & job assistance with top firms.