2018/09/25 by Bart van Merriënboer, van Merriënboer, Bart, Dan Moldovan +3 · 1 citation
Computer Science · Engineering · Physics and Astronomy · #Advanced Multi-Objective Optimization Algorithms #Computational Physics and Python Applications #Evolutionary Algorithms and Applications #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Model Reduction and Neural Networks #Robotics and Sensor-Based Localization #Software Engineering (cs.SE)
paper · pdf · doi:10.48550/arxiv.1809.09569
openalex publication_date 2018/09/25 · openalex created_date 2019/06/27 · openalex updated_date 2026/07/28
The need to efficiently calculate first- and higher-order derivatives of\nincreasingly complex models expressed in Python has stressed or exceeded the\ncapabilities of available tools. In this work, we explore techniques from the\nfield of automatic differentiation (AD) that can give researchers expressive\npower, performance and strong usability. These include source-code\ntransformation (SCT), flexible gradient surgery, efficient in-place array\noperations, higher-order derivatives as well as mixing of forward and reverse\nmode AD. We implement and demonstrate these ideas in the Tangent software\nlibrary for Python, the first AD framework for a dynamic language that uses\nSCT.\n