Start Machine Learning With Julia: Top Julia Libraries for Machine Learning

This article was published as a part of the Data Science Blogathon.

Introduction

In this article, we are going to discuss some useful Julia libraries that are very useful for Machine learning and deep learning. So let’s get started!!

Table Of Contents

• Mocha.jl
• ScikitLearn.jl
• TensorFlow.jl
• Flux
• MLBase.jl
• Merlin.jl
• Knet.jl

Mocha.jl

It is a deep learning library for the Julia programming language and it is inspired by Caffe, a C++ framework. Mocha can be utilized to train shallow/ deep convolutional neural networks, with (optional) unsupervised pre-training utilizing (stacked) auto-encoders.

Features of Mocha

• Open Source – It is open-source and licensed under MIT.
• Compatibility – It uses the HDF5 format to store model snapshots and datasets thus making it easier to cooperate with Numpy and Matlab.
• Correctness and Modularity – It has a modular architecture
• High-level Interface
• Modular Architecture – Easy to customize, compose and extend
• Minimum dependencies – No need to install any external dependencies. Just run add[‘Mocha’] and you are good to go
• It is written in Julia
• Multiple backends – This library accompanies a GPU backend, Combining highly efficient libraries from NVIDIA like cuBLAS, cuDNN, and so forth

Installation

To install this library, just run the below command:

Pkg.add(“Mocha”)

For more information, check the official documentation Link

ScikitLearn.jl

ScikitLearn.jl is a Julia wrapper for ScikitLearn Library. It implements the famous ScikitLearn algorithms and interfaces in Julia. It has models both from the Julia ecosystem and Scikit-Learn library. This library provides easy to use interface for training and testing machine learning models. It has tools to build machine learning pipelines, evaluating models, cross-validation, and hyper-parameters tuning.

Features

• It has almost 150 Python and Julia models
• Cross-validation
• DataFrames support
• Hpyer-paramteres tuning
• Feature Unions and Pipelines

Installation

To install this library, just run the below command at the REPL:

Julia>  ] add ScikitLearn

For more information, check the official documentation Link

TensorFlow.jl

Tensorflow is a famous open-source python framework by Google to build machine and deep learning models. Tensorflow.jl is a Julia wrapper around an open-source machine learning framework Tensorflow. This wrapper can be utilized for various purposes such as fast postprocessing of calculated results, fast ingestion of data mostly for data that is in uncommon format, used for visualization, and statistics calculation that doesn’t have a canned vectorized execution.

Installation

To install this library, just run the below command:

Pkg.add(“Tensorflow”)

For more information, check the official documentation Link

Flux

It is an open-source machine learning and deep learning library in Julia. It provides a single and intuitive way to define the models, just like the simple mathematical notation. Existing Julia libraries are differentiable and can be consolidated straightforwardly into Flux models. Forefront models, for example, Neural ODEs are top of the line, and Zygote empowers overhead-free gradients.

Flux is exceptionally hackable and any part can be changed, from GPU code to custom gradients and layers. Some of the features of Flux include differentiable programming, ONNX, GPU support, and compiled eager code.

Installation

So before using Flux, you’ve to first download Julia version 1.3 or later. To download Julia, you can refer to this guide Download Julia

After Julia Installation, run the below command in the Julia REPL to download Flux:

Julia>  ] add Flux

For more information, check the official documentation of Flux

MLBase.jl

It is Julia’s package that doesn’t implement any specific machine learning algorithms, but it provides some useful tools to machine learning use cases. It has a large number of useful tools to support machine learning programs.

Some of the tools that are present in this package are given below:

• Data Manipulation
• Data Preprocessing
• Performance Evaluation( e.g. ROC-AUC)
• Model Tuning (Hyper-parameter tuning)
• Cross-validation
• Score bases classification

Installation

To install this library, just run the below command:

Pkg.add(“MLBase”)

For more information, check the official documentation Link

Merlin.jl

It is a deep learning framework written in Julia to train deep neural networks. This package is very underrated. However, this would be a slip-up as Merlin.jl is an unfathomable framework that has saved me a great deal of time on a wide range of events. Like Flux.jl, Merlin is generally lightweight and written in 100% Julia code.

Merlin will in general dominate Flux in a lot of tasks, however that doesn’t imply that it will consistently be quicker. It has an implicit GPU support with CUDA. Merlin models are easy to deploy as compared to Flux. The library aims to provide a flexible, fast, and smaller profound deep learning library for machine learning.

Installation

To install this library, just run the below command at the REPL:

Julia>  ] add Merlin

For more information, check the official documentation Link

Knet.jl

Knet is another deep learning package for Julia, with a little bend. Knet was made for use at Koc University. This is one of the best packages for beginners as it is very easy to learn. This package is better maintained as compared to other packages and libraries due to its vast community. Likewise dissimilar to Flux and Merlin, Knet is somewhat of a heavier package. This is not completely written in Julia, being written in different languages, for example, C and MATLAB running beneath the Julia code. It ought to be noted, nonetheless, that a little part of Knet is written in C and MATLAB, adding up to just about 1.8 percent.

Installation

To install this library, just run the below command at the REPL:

Julia>   ] add Knet

For more information, check the official documentation Link

Conclusion

So in this article, we have covered the top 7 Machine Learning and Deep Learning libraries in Julia in 2021. I hope you learn something from this blog and it will turn out best for your project. Thanks for reading and your patience. Good luck!

You can check my articles here: Articles

Thanks for reading this article on python libraries for image processing and for your patience. Do let me in the comment section. Share this article, it will give me the motivation to write more blogs for the data science community.

Email id: gakshay1210@gmail.com

Follow me on LinkedIn: LinkedIn

The media shown in this article on Top Machine Learning Libraries in Julia are not owned by Analytics Vidhya and is used at the Author’s discretion. 

Akshay