The thesis explores the optimization of Abstract Inlining in the Glasgow Haskell Compiler (GHC), highlighting its significance and challenges. It investigates the decision space of GHC's inliner, presenting potential improvements ...
About:
Simon Peyton Jones’ personal website.
Website:
Specializations:
Incoming Links:
Subscribe to RSS:
In his keynote talk at the Haskell Symposium, Simon Peyton Jones discusses the evolution and significance of join points as an optimization tool in functional language compilers, particularly in GHC. He reflects on eight years of ...
This paper discusses the adaptation of dual-numbers construction for reverse-mode automatic differentiation (AD) in multidimensional arrays, addressing performance issues in array programs. It introduces a new algorithm comprising...
0Bits with soul
2025-01-24 •
Computer science offers a rich new lens to explore the world, allowing for mechanized analysis and transformation of information, engaging creativity, and learning from data. The talk aims to share the joy, beauty, and creativity ...
0Functional abstract interpretation
2024-12-12 •
The text discusses the use of functional programming languages and the challenges they present, as well as the evolution of the Glasgow Haskell Compiler (GHC) into an industrial strength tool. It also presents two results of work ...
...
The paper presents an efficient implementation of reverse-mode automatic differentiation for higher-order functions, with linear run time and memory consumption. Simon Peyton Jones explains the concept of automatic differentiation...
0Triemaps that match
2021-07-01 •
...
The text discusses the challenge of missing values in machine learning and proposes VICAUSE, a new approach to handle missing value imputation and causal discovery efficiently with deep learning. The method shows improved performa...
0Hashing Modulo Alpha-Equivalence
2021-06-01 •
The text discusses a new, efficient way to hash modulo alpha-equivalence in identifying identical subtrees of a program syntax tree. The method uses a weak hash combiner at one point in the construction, resulting in O(n*(log n)^2...
The text discusses the NeurIPS 2020 Education Challenge, focusing on educational diagnostic questions and the use of data on hundreds of millions of answers to MCQs to drive automatic personalized learning. The competition aimed t...