Brief summary: We are accepting papers at the Continual Lifelong Learning Workshop at the Asian Conference on Machine Learning, this December in *hybrid* format! The physical conference will be held December 12-14th in Hyderabad, India: https://www.acml-conf.org/2022/
The deadline is 31st October (submit on OpenReview): https://continual-lifelong-learners.github.io/cfp/ Abstract: Deep-learning methods require extremely large amounts of data and computational resources, and lack human-like natural abilities to quickly adapt to their surroundings and learn continually. Continual lifelong learning methods aim to bridge this gap between humans and machines. In this workshop, our goal is to bring together Asian researchers working on this topic, and connect them to other communities in the rest of the world. Similar workshops have taken place in other Machine Learning conferences, but they have largely focused on researchers from North America and Europe. This workshop will provide networking opportunities for researchers, allowing them to collaborate and work towards solving continual lifelong learning. The workshop will focus on a broad range of topics covering many aspects of continual lifelong learning, including (but not limited to): fast adaptation, forward/backward transfer, continual reinforcement learning, skill learning, abstraction and representations for lifelong learning, and relationship to similar ideas such as multi-task learning, meta learning, curriculum learning, and active learning. Invited speakers: * Balaraman Ravindran<http://www.cse.iitm.ac.in/~ravi/> (Professor of Computer Science, IIT Madras, India) * Jonghyun Choi<https://ppolon.github.io/> (Associate Professor, Yonsei University, South Korea) * Thang D Bui<https://thangbui.github.io/> (Assistant Professor, Australian National University, Australia) * Joseph K J<https://josephkj.in/> (PhD Student, IIT Hyderabad, India) Call for Papers: Machine learning and deep learning often assume all the data is available at once and accessible whenever needed during training, which is restrictive. Ideally, we want machines to learn as flexibly as humans do: humans can adapt quickly to new environments and can continue to learn throughout our lives. This is currently not possible in machine learning. Over recent years, there has been growing interest in developing systems that can adapt quickly. In Continual Lifelong Learning, methods can ideally handle a stream of incoming data from an ever-changing source, where revisiting data is challenging or even impossible. Ideally, such a system should be able to * quickly adapt to changes, * remember and faithfully transfer old knowledge to new situations, * acquire new skills but continue to do so without forgetting, * adjust to drifts in data and/or tasks, * adapt the model/architecture accordingly, and so on. Despite recent advances, many challenges remain. Different studies often formalise the problem differently and use different benchmarks. Even when there are empirical successes, there is little theoretical understanding. The field of continual lifelong learning remains an important, yet challenging, problem that we hope to discuss in this workshop. The workshop will welcome submissions from a wide variety of topics aiming to address such challenges. We invite submissions (up to 5 pages, excluding references and appendix) in the ACML 2022 format. The submission deadline is October 31st (AoE). All submissions will be managed through OpenReview: https://openreview.net/group?id=ACML.org/2022/Workshop/CLL. The review process is double-blind so submissions should be anonymised. Please edit the ACML template so that the Editors section is empty/blank. Accepted work will be presented as posters during the workshop, and select contributions will be invited to give spotlight talks during the workshop. We encourage submissions on the following topics, including but not limited to: * Fast adaptation, * Forward/backward transfer, * Continual Reinforcement Learning, * Bayesian continual learning, * Memory-based methods for continual learning, * Theory for continual lifelong learning, * Applications of continual lifelong learning, * Skill Learning, Temporal Abstractions for Continual RL, * Unsupervised, semi-supervised and self-supervised continual learning. Workshop organisers: Siddharth Swaroop<https://siddharthswaroop.github.io/>, Martin Mundt<http://owll-lab.com/>, Khimya Khetarpal<https://kkhetarpal.github.io/>, Peter Nickl<https://team-approx-bayes.github.io/>, Lu Xu<https://x-lu.github.io/>, Emtiyaz Khan<https://emtiyaz.github.io/>. Hope to see you there (virtually or in-person), Siddharth Swaroop, postdoc at Harvard University
_______________________________________________ uai mailing list uai@engr.orst.edu https://it.engineering.oregonstate.edu/mailman/listinfo/uai
