Mixed-Precision Communication-Avoiding SGD for Generalized Linear Models on GPUs
2026-06-18T08:52:18Z•9b1d848a267090790dd946887b3c65414c8d3fd3e01f4095bddcb050706c9a2d
3D Gaussian SplattingAggregate ComputingEisenstein-Jacobi networksGPUsGaussian networks','model parallelism','runtime reconfigurationHPCIoTLLM servingPCIeRFSoCcommunication-avoiding SGDdistributed particle filteringdistributed systemsdistributed trainingfault tolerancefault-tolerant broadcastinggeneralized linear modelshardware interfaceheterogeneous GPU clustersmachine learningmixed-precisionmodel placementpixel-level communicationquantum controlspot instances
What happened
This silver document is an arXiv cs.DC feed (2026-06-18) containing multiple new papers on distributed systems, ML training/serving, and specialized hardware interfaces. Key contributions include: mixed-precision Communication-Avoiding SGD (CA-SGD) for GLMs on NVIDIA GPUs with BF16 recipes and multi-GPU speedups; a field-based Distributed Particle Filtering formulation using Aggregate Computing for adaptable IoT deployments; Splaxel, a pixel-level communication-efficient distributed training framework for 3D Gaussian Splatting; ShuntServe, a cost- and fault-tolerant LLM serving system for *het
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_dc
- Record identifier
- 9b1d848a267090790dd946887b3c65414c8d3fd3e01f4095bddcb050706c9a2d
- Enrichment time
- 2026-06-18T08:52:18Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.