*Call for papers: The First International Workshop on Continual Semi-Supervised Learning - CSSL @ IJCAI 2021*
We invite researchers to contribute to the first edition of the Continual Semi-Supervised Learning workshop to be held in conjunction with IJCAI 2021, by submitting their work to the paper track of the event by May 31 2021. https://sites.google.com/view/sscl-workshop-ijcai-2021/ *Aim of the Workshop* Whereas continual learning has recently attracted much attention in the machine learning community, the focus has been mainly on preventing the model updated in the light of new data from ‘catastrophically forgetting’ its initial knowledge and abilities. This, however, is in stark contrast with common real-world situations in which an initial model is trained using limited data, only to be later deployed without any additional supervision. In these scenarios the goal is for the model to be incrementally updated using the new (unlabelled) data, in order to adapt to a target domain continually shifting over time. The aim of this workshop is to formalise this new *semi-supervised continual learning* paradigm, and to introduce it to the machine learning community in order to mobilise effort in this direction. We will present the first two benchmark datasets for this problem, derived from important computer vision scenarios, and propose the first *Continual Semi-Supervised Learning Challenges* to the research community. *Continual semi-supervised learning* In continual semi-supervised learning, an initial training batch of data-points annotated with ground truth (class labels for classification problems, or vectors of target values for regression ones) is available and can be used to train an initial model. Then, however, the model is incrementally updated by exploiting the information provided by a time series of *unlabelled* data points, each of which is generated by a data generating process (modelled, as typically assumed, by a probability distribution) which varies with time. No artificial subdivision into ‘tasks’ is assumed, as the data-generating distribution may arbitrary vary over time. *Topics of the workshop* The goal of this workshop is to propose to the research community in artificial intelligence and machine learning the new continual semi-supervised learning problem. At the same time, we will accept papers on continual learning in its broader sense, covering for instance the following topics: · Analysis of suitability of existing datasets for continual learning. · New benchmark datasets explicitly designed for continual learning settings. · Protocols for training and testing in different continual learning settings. · Metrics for assessing continual learning methods. · Task-based continual learning. · Relation between continual learning and model adaptation. · Learning of new classes as opposed to learning from new instances. · Real-world applications of continual learning. · Catastrophic forgetting and mitigation strategies. · Applications of transfer learning, multi-task and meta-learning to continual learning. · Continual supervised, semi-supervised and unsupervised learning. · Lifelong, few-shot learning. · Continual reinforcement and inverse reinforcement learning. The list is in no way exhaustive, as the aim is to foster the debate around all aspects of continual learning, especially those which are subject of ongoing frontier research. We will invite both paper track contributions on these topics, as well as submissions of entries to a set of challenges specifically designed to test CSSL approaches. Two benchmarks will be introduced which are specifically designed to assess continual semi-supervised learning on two important computer vision tasks: *activity recognition* and *crowd counting*. A separate Call for Participation in the Challenges will be issued shortly. *Workshop format* The workshop will be a full-day event, articulated into: · Introduction by the organisers. · Presentation of the new benchmark datasets and associated Challenges. · Invited talks by top researchers in the area, with brief Q&A session after each invited talk. · Oral presentations for the Best Paper and the Best Student Paper. · Spotlight talks for the winners of the Challenges. · Two poster sessions (a morning session and an afternoon session) for all other accepted papers. · Discussion panel on the future of continual learning from long streams of unlabelled data. There will be 5 invited talks, with a projected duration of 9 hours and 30 minutes, including 1 hour and 10 minutes for breaks. *Important dates* Paper submission: May 31 2021 Notification: June 30 2021 Camera-ready: July 31 2021 Workshop date: August 21-23 2021 (to be confirmed) *Submission guidelines* Papers submitted to the workshop will follow the standard IJCAI 2021 template (6 pages plus 1 for the references), see https://www.ijcai.org/authors_kit Paper submission will take place through EasyChair at: https://easychair.org/my/conference?conf=csslijcai2021 The organisers are negotiating with top publishers the nature of the proceedings, further details will be provided soon. Authors are welcome to submit a supplementary material document with details on their implementation; however, reviewers are not required to consult this additional material when assessing the submission. *The Workshop will allow for the submission of papers concurrently submitted elsewhere, with the aim of aggregating all relevant efforts in this area*. *Double-blind review: *Authors must not include any identifying information (names, affiliations, etc.) or links and self-references that may reveal their identities. The organisers aim to provide feedback from three reviewers per submission, which will assess the submission based on relevance, novelty and potential for impact. Reviewers are asked to assess the submission (Reject/Borderline/Accept) as well as provide written feedback. There will be no additional rebuttal period. The authors of accepted papers must guarantee their presence at the workshop. At least one author for each accepted paper must register for the conference. The same holds for Challenge winners. *Awards* The Workshop will issue: · A Best Paper Award to the author(s) of the best accepted paper, as judged by the Organising Committee based on the reviews assigned by PC members. · A Best Student Paper Award, selected in the same way. · A Prize to be awarded to the winners of each of the Challenges. We reserve the right to issue Honourable Mentions to the most original challenge entries. *Invited speakers* · Razvan Pascanu (Deepmind) · Tinne Tuytelaars (KU Leuven) · Chelsea Finn (Stanford) · Bing Liu (University of Illinois at Chicago) *Organising committee* · Fabio Cuzzolin (Oxford Brookes University) · Kevin Cannons (Huawei Technologies Canada) · Vincenzo Lomonaco (University of Pisa and ContinualAI) · Irina Rish (University of Montreal and MILA) · Salman Khan (Oxford Brookes University) · Mohamad Asiful Hossain (Huawei Technologies Canada) · Ajmal Shahbaz (Oxford Brookes University)
_______________________________________________ uai mailing list uai@engr.orst.edu https://it.engineering.oregonstate.edu/mailman/listinfo/uai
