( events)   Timezone: »  
Mon Dec 13 09:00 AM -- 05:55 PM (PST)
ML For Systems
Benoit Steiner · Jonathan Raiman · Martin Maas · Azade Nazi · Mimee Xu · Anna Goldie

Workshop Home Page

ML for Systems is an emerging research area that has shown promising results in the past few years. Recent work has shown that ML can be used to replace heuristics, solve complex optimization problems, and improve modeling and forecasting when applied in the context of computer systems.

As an emerging area, ML for Systems is still in the process of defining the common problems, frameworks and approaches to solving its problems, which requires venues that bring together researchers and practitioners from both the systems and machine learning communities. Past iterations of the workshops focused on providing such a venue and broke new ground on a broad range of emerging new directions in ML for Systems. We want to carry this momentum forward by encouraging the community to explore areas that have previously received less attention. Specifically, the workshop commits to highlighting works that also optimize for security and privacy, as opposed to metrics like speed and memory and use ML to optimize for energy usage and carbon impact. Additionally, this year we will encourage the development of shared methodology, tools, and frameworks.

For the first time since the inception of the workshop, we will organize a competition. This competition will showcase important systems problems, and challenges the ML community to test their methods and algorithms on these problems. Our competition tasks are designed to have a low barrier of entry that attracts newcomers as well as systems veterans.

This setup will allow attendees to meet with top researchers and domain experts, old and new, bridging cutting edge ML research with practical systems design. We hope that providing a prestigious venue for researchers from both fields to meet and interact will result in both fundamental ML research as well as real-world impact to computer systems design and implementation.

Opening Remarks (Introduction)
Learned Systems (Invited Talk)
Accelerating Systems and ML for Science (Invited Talk)
Engineering approximate computations (Invited Talk)
Lunch Break (Break)
Learned Compiler Optimizations (Invited Talk)
ML for Autotuning Production ML Compilers (Invited Talk)
ML-guided iterative refinement for system optimization (Invited Talk)
Closing Remarks (Outro)
Community Infrastructure for Applying Reinforcement Learning to Compiler Optimizations (Spotlight)
Resource Allocation in Disaggregated Data Centre Systems with Reinforcement Learning (Spotlight)
Data-Driven Offline Optimization for Architecting Hardware Accelerators (Spotlight)
Low-Precision Training in Logarithmic Number System using Multiplicative Weight Update (Spotlight)
Neuroevolution-Enhanced Multi-Objective Optimization for Mixed-Precision Quantization (Spotlight)
DeepRNG: Towards Deep Reinforcement Learning-Assisted Generative Testing of Software (Spotlight)
Achieving Low Complexity Neural Decoders via Iterative Pruning (Spotlight)
Automap: Towards Ergonomic Automated Parallelism for ML Models (Spotlight)
Learning to Combine Instructions in LLVM Compiler (Spotlight)
Towards Intelligent Load Balancing in Data Centers (Spotlight)
Reinforced Workload Distribution Fairness (Spotlight)
Interpretability of Machine Learning in Computer Systems: Analyzing a Caching Model (Spotlight)
Generative Optimization Networks for Memory Efficient Data Generation (Spotlight)