Call for Papers

Important Dates

Paper submissions due: Friday, March 31, 2023, 11:59 PM AoE

Notification to authors: Tuesday, April 18, 2023

Overview

Using ML for improving computer systems has seen a significant amount of work both in academia and industry. However, deployed uses of such techniques remain rare. While many published works in this space focus on solving the underlying learning problems, we observe a gap between these fundamental learning methods and actually using these methods to drive real world impact. The goal of this workshop is to raise awareness of this gap and to seek understanding of why it occurs. Examples of topics include: Highlighting potential issues with ML feature stability, reliability, availability, ML integration into rollout processes, verification, safety guarantees, feedback loops introduced by learning, debuggability, and explainability. 

We hope to identify these problems and bring together practitioners and academic researchers, both on the production systems and ML side, to work towards a methodology for capturing these problems in academic research. We believe that starting this conversation between the academic and industrial research communities will facilitate the adoption of ML for Systems research in production systems, and will provide the academic community with access to new research problems that exist in real-world deployments but have seen less attention in the academic community.

To this end, we invite lightweight submissions (between 1-4 pages, excluding references and appendices) in the broad area of challenges associated with using machine learning in computer systems. We encourage a wide variety of submissions, from success-stories to retrospectives, and are particularly interested in anecdotes and experiences from real-world deployments. We also encourage academic and research submissions that are describing techniques that could be applicable to such deployments, even if they do not describe a real system.

Topics of Interest

Some areas of interest include:

• Research on techniques that expose, address problems with, or are otherwise related to the deployment of ML and data-driven methods in systems, particularly in (but not limited to) the following areas:

  • Feature stability, reliability, or availability

  • ML integration into rollout processes

  • Verification for ML models

  • Safety guarantees for ML models

  • Control system aspects of ML for systems, such as feedback loops

  • Debuggability and explainability

• Retrospectives, success stories, or negative results from deployments of ML techniques in computer systems.

• Descriptions of systems where ML was considered but ultimately not adopted.

• Benchmarks that provide industry & academia with the opportunity to capture aspects of deployment of ML for systems.

We define computer systems to broadly include (not limited to): Computer Architecture, Network Systems, Operating Systems, Runtime Systems, Software Systems, Code Modeling, Compilers, Databases, Data Centers, Distributed Systems, Security and Performance Tools. We define ML to broadly include any data-driven methods (not limited to deep learning).

Submission Instructions

• To prepare your submission please use the LaTeX style files provided at MLSys2023 Style. Submitted papers will be in a 2-column format and can be up to 4 pages long, not including references. All author names and affiliations should appear on the title page.

• There will be no formal proceedings for this workshop. Authors may publish their work in other journals or

conferences. The workshop will also include a series of invited talks, panels, and breakout sessions.

• All submissions should be in PDF format and they should be submitted through the submission form.

• Submission of a paper serves as an agreement that at least one of the authors will attend the workshop to present the paper. If the workshop registration fee prevents the presenter from attending, please contact chairs@pacmi-workshop.org.

Submission Site: https://pacmi23.hotcrp.com/