Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
April 25, 2026ACM SIGMOD Record

Technical Perspective on 'MEMPHIS: Holistic Lineage-based Reuse and Memory Management for Multi-backend ML Systems'

View Full Paper
Ask AI
Bookmark
Share

Authors

AKArun Kumar

Discussion

Loading...

Member takes

Overview

Technical perspective explores managing memory and computation in heterogeneous ML systems, highlighting challenges.

Key Points

  • This research addresses the challenge of managing computation reuse and memory across heterogeneous machine learning backends with different models.
  • Analyzed the execution models of various ML platforms including local CPUs, GPUs, and Apache Spark.
  • Evaluated the memory hierarchies and caching mechanisms associated with these platforms.
  • Proposed a holistic approach to streamline memory and computation management across diverse backends.
  • Identified gaps in current research addressing multi-backend ML systems.
  • Demonstrated that current methods for memory management are insufficient for heterogeneous environments.
  • Outlined key strategies for enhancing management efficiency across different execution contexts.

Cite This Study

Arun Kumar (2026) studied this question.

synapsesocial.com/papers/69ec5a6b88ba6daa22dabfcbhttps://doi.org/10.1145/3810900.3810915
View Full Paper
Ask AI
Bookmark
Share