Skip to main content
UMass Collegiate M The University of Massachusetts Amherst
  • Visit
  • Apply
  • Give
  • Search UMass.edu
Manning College of Information & Computer Sciences

Main navigation

  • Academics

    Programs

    Undergraduate Programs Master's Programs Doctoral Program Graduate Certificate Programs

    Academic Support

    Advising Career Development Academic Policies Courses Scholarships and Fellowships
  • Research

    Research

    Research Areas Research Centers & Labs Undergraduate Research Opportunities

    Faculty & Researchers

    Faculty Directory Faculty Achievements Turing Award

    Engage

    Research News Distinguished Lecturer Series Rising Stars in Computer Science Lecture Series
  • Community

    On-Campus

    Diversity and Inclusion Student Organizations Massenberg Summer STEM Program Awards Programs Senior Celebration

    External

    Alumni Support CICS
  • People
    Full A-Z Directory Faculty Staff
  • About

    Overview

    College Overview Leadership Our New Building

    News & Events

    News & Stories Events Calendar

    Connect

    Visiting CICS Contact Us Employment Offices & Services
  • Info For
    Current Undergraduate Students Current Graduate Students Faculty and Staff Newly Accepted Undergraduate Students

PhD Dissertation Proposal Defense: Sandeep Polisetty, Abstractions to Eliminate Redundancy in Training Graph Neural Networks on GPUs

Content

Monday, June 30, 2025, 10:00 AM - Monday, June 30, 2025, 12:00 PM

Online
PhD Dissertation Proposal Defense
Presentation

Speaker

Sandeep Polisetty

Abstract

Graph machine learning has achieved significant success in extending deep learning to graph-structured data, removing the need for hand-engineered features. These models typically use message passing across edges—expressed as sparse computations—to capture the relational structure inherent in graphs. However, current abstractions for mapping such computations onto highly parallel processors like GPUs are often inefficient, introducing redundancy due to the dense and irregular connectivity of graph data.

In the first part of my thesis, I introduce split parallelism, a novel  
abstraction that addresses the limitations of traditional data parallelism on GPUs. In this approach, each node is assigned to a specific GPU, and information from remote neighbors is communicated across GPUs at every hidden layer. We also present an API that allows various graph machine learning models to be expressed with minimal modification to the standard training loop. To further optimize this approach, we propose a graph partitioning strategy that minimizing inter-GPU communication for split-parallel training. Our implementation of split parallelism achieves up to 1.7× speedup over state-of-the-art methods.

We find that one drawback of split-parallelism as we attempt to train  
larger/deeper models is that communication and synchronization costs can become substantive. For future work, I aim to reduce these costs by exploring hybrid forms of parallelism:

  • Pipeline + Split Parallelism: This approach introduces controlled  
    redundancy across multiple batches, overlapping computation of one batch’s split with the communication of another, thereby effectively hiding communication costs.
  • Data + Split Parallelism: Here, split parallelism is selectively applied  
    only to certain layers, while other layers use traditional data parallelism, reducing overhead of communication and synchronization.

Advisors

Hui Guan and Marco Serafini


Host Name: Sandeep Polisetty
Host URL: https://umass-amherst.zoom.us/j/99105166451
 

Online event posted in PhD Dissertation Proposal Defense

More link

Join via Zoom

Site footer

Manning College of Information & Computer Sciences
  • Find us on Facebook
  • Find us on YouTube
  • Find us on LinkedIn
  • Find us on Instagram
  • Find us on Flickr
  • Find us on Bluesky Social
Address

140 Governors Dr
Amherst, MA 01003
United States

  • Visit CICS
  • Give
  • Contact Us
  • Employment
  • Events Calendar
  • Offices & Services

Info For

  • Current Undergraduate Students
  • Current Graduate Students
  • Faculty & Staff
  • Newly Accepted Undergraduate Students

Global footer

  • ©2025 University of Massachusetts Amherst
  • Site policies
  • Privacy
  • Non-discrimination notice
  • Accessibility
  • Terms of use