Call for Papers

Topics of Interest

This meetup is intended to provide an opportunity for connecting with researchers interested in the efficiency of machine learning training and inference. We welcome posters on any research related to these topics, including applied work, empirical analysis, theoretical research, industrial experience, and meta-analysis.

You are welcome to submit posters on work-in-progress and papers/publications that have been released within the past year.

Please make sure that all authors who will present the poster on site also register to attend the event!

Example topics of interest include:

• Sparsity

  • Algorithms for sparsity: pruning / quantization, scaling laws for sparsity, sparse training

  • Theory and science of sparsity: When is overparameterization necessary (or not), representation ability of sparse networks, sparsity and generalization, stability of sparse models, forgetting due to sparsity, fairness and privacy concerns

  • Applications for sparsity: Resource-efficient and data-efficient learning at the edge or in the cloud, communication-efficient distributed and federated learning with sparse models, graph and network science applications

  • Augmentation

  • Data augmentation schemes, how to fit in-distribution or to generalize out-of-distribution, affinity and diversity

  • Model augmentation techniques and model soups, label augmentation

  • Algorithmic training speed up

    1. • Last-layer retraining, minimizing the cost of backpropagation, model update schemes and model patching

  • Architecture search

    1. • Efficient architectures, evolutionary algorithms, hyperparameter optimization

       • Meta-learning and transfer learning

Important Dates

Submissions Open: May 10th, 2022

Submissions Due: June 7th, 2022 CET

Author Notification: June 9th, 2022

Workshop Date: June 13th, 2022

Submission Instructions

The workshop submission and review will be handled by OpenReview: https://openreview.net/group?id=EfficientML/2022/Workshop
We expect to receive a short abstract describing your work and (advisable but not absolutely necessary) a PDF supporting your submission (it could be a recently accepted paper or a preprint).
Poster submission (accepted abstracts): Once your submission is accepted, we will ask you to bring a physical poster to the event in the A0/A1 format.

Posters will also be put online. They should be in the landscape format and must be a PNG file sent to us per email. The maximum dimensions of a poster are 5120 x 2880 pixels and no more than 10 MB. Thumbnails must be a PNG file and should be no larger than 320 x 256 pixels and no more than 5MB. You can use templates here.

Criteria for Acceptance

Our goal is to give the community a chance to reconnect after two years of virtual events. As such, we intend to accept all submissions that are (1) relevant to the topic area EfficientML, (2) technically well-substantied, and (3) non-trivial and previously unknown results.

Reviewing will be conducted in a double-blind fashion. We will also select two outstanding posters to highlight.