Divide-and-conquer checkpointing for arbitrary programs with no user annotation

JM Siskind, BA Pearlmutter - Optimization Methods and Software, 2018 - Taylor & Francis
Classical reverse-mode automatic differentiation (AD) imposes only a small constant-factor
overhead in operation count over the original computation, but has storage requirements …

Efficient implementation of a higher-order language with built-in AD

JM Siskind, BA Pearlmutter - arxiv preprint arxiv:1611.03416, 2016 - arxiv.org
We show that Automatic Differentiation (AD) operators can be provided in a dynamic
language without sacrificing numeric performance. To achieve this, general forward and …

Mini-symposium on automatic differentiation and its applications in the financial industry

S Geeraert, CA Lehalle, BA Pearlmutter… - ESAIM: Proceedings …, 2017 - esaim-proc.org
Automatic differentiation has been involved for long in applied mathematics as an alternative
to finite difference to improve the accuracy of numerical computation of derivatives. Each …

AD in Fortran, Part 1: Design

A Radul, BA Pearlmutter, JM Siskind - arxiv preprint arxiv:1203.1448, 2012 - arxiv.org
We propose extensions to Fortran which integrate forward and reverse Automatic
Differentiation (AD) directly into the programming model. Irrespective of implementation …

[PDF][PDF] Pearlmutter & Siskind joint work on Automatic Differentiation (AD)

BA PEARLMUTTER - engineering.purdue.edu
AG Baydin, BA Pearlmutter, AA Radul, and JM Siskind. 2015b. Automatic differentiation in
machine learning: a survey. Technical Report 1502.05767. arxiv. http://arxiv …

[PDF][PDF] The Differentiable Curry

MARWG Plotkin, DVD Belov - dimitriv.github.io
Differentiable programming allows programmers to calculate program gradients and unlocks
experimentation with new optimizers and neural network architectures. This is why modern …

[PDF][PDF] AD in Fortran

A Radul, BA Pearlmutter, JM Siskind - arxiv preprint arxiv:1203.1448, 2012 - core.ac.uk
We propose extensions to FORTRAN which integrate forward and reverse Automatic
Differentiation (AD) directly into the programming model. Irrespective of implementation …

[PDF][PDF] Thesis title: The Culprit Pointer Method for Selective Backtracking Advisor: David A. McAllester Massachusetts Institute of Technology Ph. D. in computer …

JM Siskind - Consultant, 1987 - engineering.purdue.edu
MetaLogic, Inc. Cambridge, Massachusetts April 1982–August 1985 Founder, Chairman,
Treasurer, Vice President, and CTO Founded MetaLogic to commercialize the MACPITTS …