The ArrayFire Blog

Researchers from the University Bordeaux in France credit ArrayFire in a paper published in ICPR 2020’s workshop on Explainable Deep Learning for AI. The paper is titled “Samples Classification Analysis Across DNN Layers with Fractal Curves.” It provides a tool for visualizing where the deep neural network starts to be able to discriminate the classes. Summary Deep neural networks (DNN) are becoming the prominent solution when using machine learning models. However, they suffer from a black-box effect that complicates their inner workings interpretation and thus the understanding of their successes and failures. Information visualization is one way, among others, to help in their interpretability and hypothesis deduction. This paper presents a novel way to visualize atrained DNN to depict at the same …

Researchers from the University of Central Florida credit ArrayFire in a paper published in IEEE Transactions on Signal Processing. The paper is titled “An Exact and Fast Computation of Discrete Fourier Transform for Polar and Spherical Grid” and provides the first exact and fast solution to the problem of obtaining discrete Fourier transform for polar and spherical grids. This paper is fully reproducible on Github. Summary Numerous applied problems of two-dimensional (2-D) and 3-D imaging are formulated in the continuous domain. They emphasize obtaining and manipulating the Fourier transform in polar and spherical coordinates. However, translating continuum ideas with discretely sampled data on a Cartesian grid is problematic. There exists no exact and fast solution to the problem of obtaining discrete Fourier …

Researchers from the Catalan Telecommunications Technology Centre in Spain credit ArrayFire in a paper published in the Applied Soft Computing Journal. The paper is titled “A GPU fully vectorized approach to accelerate performance of NSGA-2 based on stochastic non-domination sorting and grid-crowding” and showcases ArrayFire accelerating decision space exploration for multi-objective optimization problems. Summary This work introduces an accelerated implementation of NSGA-2 on a graphics processing unit (GPU) to reduce execution time. Parallelism is achieved at the population level using vectorization. All the algorithm components are run on the device, minimizing communication overhead. New stochastic versions of both non-domination sorting and crowding are introduced in the article. They are designed to be efficiently vectorized on GPU; therefore, the proposed approach is finally …