Análise do potencial da linguagem Elixir no setor de redes neurais

Abstract

With a wide range of libraries focused on the machine learning market, such as TensorFlow, NumPy, Pandas, Keras, and others, Python has made its name as one of the leading languages for this area of programming. In February 2021, Jose Valim and Sean Moriarity published the first version of the Numerical Elixir (Nx) library, a library for tensor operations written in the functional programming language Elixir created by Brazilian Jose Valim. The Nx library aims to enable the language to be a great choice for operations computationally intensive in the use of graphics processing units. This work compares the results of Python and Elixir in training convolutional neural networks using MNIST and CIFAR-10 datasets, concluding that Python achieved better results overall, having 22.39% shorter training time than Elixir for the same scenarios. On the other hand, this work also concludes that Elixir is already a viable alternative for this sector, since it achieved good average performance, better optimization in the use of random access memory (RAM) and proved to be a good alternative for those working in the functional languages industry.