<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<b class=""><u class="">ScaDL 2021: Third IPDPS Workshop on Scalable Deep Learning over Parallel and Distributed Infrastructure</u></b><br class="">
<a href="https://2021.scadl.org/" class="">https://2021.scadl.org</a>
<div class=""><a href="https://2021.scadl.org/call-for-papers" class="">https://2021.scadl.org/call-for-papers</a><br class="">
<div class=""><br class="">
<b class="">Scope of the Workshop</b><br class="">
Recently, Deep Learning (DL) has received tremendous attention in the research community because of the impressive results obtained for a large number of machine learning problems. The success of state-of-the-art deep learning systems relies on training deep
neural networks over a massive amount of training data, which typically requires a large-scale distributed computing infrastructure to run. In order to run these jobs in a scalable and efficient manner, on cloud infrastructure or dedicated HPC systems, several
interesting research topics have emerged which are specific to DL. The sheer size and complexity of deep learning models when trained over a large amount of data makes them harder to converge in a reasonable amount of time. It demands advancement along multiple
research directions such as, model/data parallelism, model/data compression, distributed optimization algorithms for DL convergence, synchronization strategies, efficient communication and specific hardware acceleration.<br class="">
<br class="">
<b class="">SCADL seeks to advance the following research directions:</b><br class="">
<ul class="MailOutline">
<li class=""><u class="">Asynchronous and Communication-Efficient SGD:</u> Stochastic gradient descent is at the core of large-scale machine learning. Parallelizing SGD gradient computation across multiple nodes increases the data processed per iteration, but
exposes the SGD to communication and synchronization delays and unpredictable node failures in the system. Thus, there is a critical need to design robust and scalable distributed SGD methods to achieve fast error-convergence in spite of such system variabilities.</li><li class=""><u class="">High performance computing aspects:</u> Deep learning is highly compute intensive. Algorithms for kernel computations on commonly used accelerators (e.g. GPUs), efficient techniques for communicating gradients and loading data from
storage are critical for training performance.</li><li class=""><u class="">Model and Gradient Compression Techniques:</u> Techniques such as reducing weights and the size of weight tensors help in reducing the compute complexity. Using lower-bit representations allow for more optimal use of memory and communication
bandwidth.</li></ul>
<br class="">
This intersection of distributed/parallel computing and deep learning is becoming critical and demands specific attention to address the above topics which some of the broader forums may not be able to provide. The aim of this workshop is to foster collaboration
among researchers from distributed/parallel computing and deep learning communities to share the relevant topics as well as results of the current approaches lying at the intersection of these areas.<br class="">
<br class="">
<b class="">Areas of Interest</b><br class="">
In this workshop, we solicit research papers focused on distributed deep learning aiming to achieve efficiency and scalability for deep learning jobs over distributed and parallel systems. Papers focusing both on algorithms as well as systems are welcome. We
invite authors to submit papers on topics including but not limited to: </div>
<div class=""><br class="">
Deep learning on cloud platforms, HPC systems, and edge devices<br class="">
- Model-parallel and data-parallel techniques<br class="">
- Asynchronous SGD for Training DNNs<br class="">
- Communication-Efficient Training of DNNs<br class="">
- Scalable and distributed graph neural networks Sampling techniques for graph neural networks<br class="">
- Federated deep learning, both horizontal and vertical, and its challenges<br class="">
- Model/data/gradient compression<br class="">
- Learning in Resource constrained environments<br class="">
- Coding Techniques for Straggler Mitigation<br class="">
- Elasticity for deep learning jobs/spot market enablement<br class="">
- Hyper-parameter tuning for deep learning jobs<br class="">
- Hardware Acceleration for Deep Learning<br class="">
- Scalability of deep learning jobs on large clusters<br class="">
- Deep learning on heterogeneous infrastructure<br class="">
- Efficient and Scalable Inference<br class="">
- Data storage/access in shared networks for deep learning<br class="">
<br class="">
<b class="">Format</b><br class="">
Due to the continuing impact of COVID-19, ScaDL 2021 will also adopt relevant IPDPS 2021 policies on virtual participation and presentation. Consequently, the organizers are currently planning a hybrid (in-person and virtual) event.</div>
<div class=""><br class="">
</div>
<div class=""><b class="">Submission Link</b><br class="">
Please log in to Linklings using <a href="https://www.google.com/url?q=https://ssl.linklings.net/conferences/ipdps/&sa=D&sntz=1&usg=AFQjCNFDxMQiNuFK5eHp4M0CyZF-qIKV1w" class="">this link</a> (create an account if necessary). Once you login, you will find a
link to submissions for the ScaDL workshop.<br class="">
<br class="">
<b class="">Key Dates</b><br class="">
<b class="">Paper Submission: February 1, 2021</b><br class="">
Acceptance Notification: March 15, 2021<br class="">
Camera-ready due: March 30, 2021<br class="">
Workshop: May 21, 2021<br class="">
<br class="">
<b class="">Author Instructions</b><br class="">
ScaDL 2021 accepts submissions in three categories:<br class="">
- Regular papers: 8-10 pages<br class="">
- Short papers: 4 pages<br class="">
- Extended abstracts: 1 page<br class="">
The aforementioned lengths include all technical content, references and appendices.<br class="">
Papers should be formatted using IEEE conference style, including figures, tables, and references. The IEEE conference style templates for MS Word and LaTeX provided by IEEE eXpress Conference Publishing are available for download. See the latest versions at <a href="https://www.ieee.org/conferences/publishing/templates.html" class="">https://www.ieee.org/conferences/publishing/templates.html</a></div>
<div class=""><br class="">
<b class="">General Chairs</b><br class="">
Stacy Patterson, RPI, USA</div>
<div class="">Parijat Dube, IBM Research, USA<br class="">
<br class="">
<b class="">Program Committee Chairs</b><br class="">
Yogish Sabharwal, IBM Research, India</div>
<div class="">Danilo Ardagna, Politecnico di Milano, Italy<br class="">
<br class="">
<b class="">Logistics & Web Chair</b><br class="">
Jayaram K. R., IBM Research, USA<br class="">
<br class="">
<b class="">Publicity Chairs</b><br class="">
Federica Filippini, Politecnico di Milano, Italy</div>
<div class="">Anirban Das, RPI, USA<br class="">
<br class="">
<b class="">Program Committee</b><br class="">
See the workshop website <a href="https://2021.scadl.org/" class="">https://2021.scadl.org</a></div>
<div class=""><br class="">
<b class="">Steering Committee</b></div>
<div class="">Vijay K. Garg, University of Texas at Austin</div>
<div class="">Vinod Muthusamy, IBM Research AI<br class="">
Ashish Verma, IBM Research AI</div>
</div>
</body>
</html>