Mixed-Precision Communication-Avoiding SGD for Generalized Linear Models on GPUs (Q7266589)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
No description defined
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Mixed-Precision Communication-Avoiding SGD for Generalized Linear Models on GPUs |
No description defined |
Statements
Mixed-Precision Communication-Avoiding SGD for Generalized Linear Models on GPUs (English)
0 references
16 June 2026
0 references
cs.DC
0 references
cs.LG
0 references
math.NA
0 references
stat.ML
0 references
Aditya Devarakonda
0 references
Irene Simo Munoz
0 references
Giulia Guidi
0 references
This paper studies mixed-precision communication-avoiding SGD for generalized linear models on NVIDIA GPUs, decomposing rounding error into nine independent precision choices tied only to hardware unit roundoffs. On NERSC Perlmutter A100 GPUs, the method matches FP32 SGD loss within 0.5% while achieving 5.1-6.8x speedup.
0 references
Stochastic Gradient Descent
0 references
Mixed Precision
0 references
GPU Computing
0 references
Communication-Avoiding Algorithms
0 references
Distributed Optimization
0 references
