Intézeti szeminárium

Interval Arithmetic on ML Infrastructure: A Path Toward Fast Interval CSP

Interval arithmetic is essential in many scientific and engineering fields, but its adoption is often hindered by computational complexity. Specialized software tools are commonly used to address this challenge. In this work, we demonstrate that, with targeted modifications, the widely used deep learning frameworks such as TensorFlow can serve as fast and reliable platforms for interval computations. Our approach leverages TensorFlow's powerful infrastructure for parallelization, optimization, and hardware acceleration. We present the necessary modifications, discuss key algorithms and methods employed, and generalize our findings. This work lays the foundation for further developments, including the application of ML frameworks to solving interval constraint satisfaction problems (CSPs).