[NeurIPS workshop][CfP] All things Attention: Bridging different perspectives on attention

[Abhijat Biswas]

Hi Everyone,

On behalf of the co-organisers, we would like to invite you to submit your work to our NeurIPS workshop on “All Things Attention: Bridging Different Perspectives on Attention”. The details of the workshop and submission instructions are as follows:

All Things Attention: Bridging Different Perspectives on Attention

The Thirty Sixth Conference on Neural Information Processing Systems (NeurIPS)

Dec 2, 2022

NeurIPS 2022 is a hybrid Conference

https://attention-learning-workshop.github.io/

The All Things Attention workshop aims to foster connections across disparate academic communities that conceptualize "Attention" such as Neuroscience, Psychology, Machine Learning, and Human Computer Interaction. Workshop topics of interest include (but are not limited to):

1. Relationships between biological and artificial attention

   1. What are the connections between different forms of attention in the human brain and present deep neural network architectures? 

   2. Can the anatomy of human attention models provide usable insights to researchers designing architectures for artificial systems? 

   3. Given the same task and learning objective, do machines learn attention mechanisms that are different from humans? 

1. Attention for reinforcement learning and decision making

   1. How have reinforcement learning agents leveraged attention in decision making?

   2. Do decision-making agents today have implicit or explicit formalisms of attention?

   3. How can AI agents build notions of attention without explicitly baked in notions of attention?

   4. Can attention significantly enable AI agents to scale e.g. through gains in sample efficiency, and generalization?

   5. How should learning systems reason about computational attention (which parts of sensed inputs to focus computation on)?

2. Benefits and formulation of attention mechanisms for continual / lifelong learning

   1. How can continual learning agents optimize for retention of knowledge for tasks that it already learned? 

   2. How can the amount of interference between different inputs be controlled via attention? 

   3. How does the executive control of attention evolve with learning in humans? 

   4. How can we study the development of attentional systems in infancy and childhood to better understand how attention can be learned?

3. Attention as a tool for interpretation and explanation

   1. How have researchers leveraged attention as a visualization tool?

   2. What are the common approaches when using attention as a tool for interpretability in AI? 

   3. What are the major bottlenecks and common pitfalls in leveraging attention as a key tool for explaining the decisions of AI agents?

   4. How can we do better?

4. The role of attention in human-computer interaction and human-robot interaction

   1. How do we detect aspects of human attention during interactions, from sensing to processing to representations?   

   2. What systems benefit from human attention modeling, and how do they use these models?

   3. How can systems influence a user’s attention, and what systems benefit from this capability?

   4. How can a system communicate or simulate its own attention (humanlike or algorithmic) in an interaction, and to what benefit?

   5. How do attention models affect different applications, like collaboration or assistance, in different domains, like autonomous vehicles and driver assistance systems, learning from demonstration, joint attention in collaborative tasks, social interaction, etc.?

   6. How should researchers thinking about attention in different biological and computational fields organize the collection of human gaze data sets, modeling gaze behaviors, and utilizing gaze information in various applications for knowledge transfer and cross-pollination of ideas?

5. Attention mechanisms in Deep Neural Network (DNN) architectures

   1. How does attention in DNN such as transformers relate to existing formalisms of attention in cogsci/psychology? 

   2. Do we have a concrete understanding of how and if self-attention in transformers contributes to its vast success in recent models such as GPT2, GPT3, DALLE.? 

   3. Can our understanding of attention from other fields inform the progress we have achieved in recent breakthroughs?

SUBMISSION INSTRUCTIONS

We invite you to submit papers (up to 9 pages for long papers and up to 5 pages for short papers, excluding references and appendix) in the NeurIPS 2022 format. All submissions will be managed through OpenReview (submission website). Supplementary Materials uploads are to only be used optionally for extra videos/code/data/figures and should be uploaded separately in the submission website.

The review process is double-blind so the submission should be anonymized. Accepted work will be presented as posters during the workshop, and select contributions will be invited to give spotlight talks during the workshop. Each accepted work entering the poster sessions will have an accompanying pre-recorded 5-minute video. Please note that at least one coauthor of each accepted paper will be expected to have a NeurIPS conference registration and participate in one of the poster sessions. 

Submissions will be evaluated based on novelty, rigor, and relevance to the theme of the workshop. Both empirical and theoretical contributions are welcome. Submissions should not have previously appeared in a journal or conference (including accepted papers to NeurIPS 2022). Submissions must adhere to the NeurIPS Code of Conduct.