Leaving 🌧️ SF for 🌞 New Orleans. Excited to go to #NeurIPS2022 for the first time! You can catch me Friday at the AllThingsAttention workshop, and Saturday at the Gaze meets ML workshop (gaze-meets-ml.github.io/gaze_ml_2022/). Looking forward to seeing you there! 👀

Decomposing tasks based on independent factors (like objects) leads to combinatorial generalization in object rearrangement problems! Check out Michael Chang's fantastic work at his oral in the #NeurIPS2022 Attention workshop and the thread below 👇

Affordances can be characterized as hard attention over temporally-extended choices in RL. Using affordances in RL can be understood through the paradox of choice - fewer but more meaningful choices are better. Joint work led by Andrei Nica at #NeurIPS2022 Attention workshop 🧵

After a productive first poster session we are on to the second half of the workshop, with invited and spotlight speakers! #NeurIPS2022

Enjoyed being a part of the panel discussion today at #NeurIPS2022 at the All Things Attention workshop. We discussed a number of issues pertaining to attention 👇

How can we use attention for continual learning? How can we make attention more interpretable? ... Lots of interesting perspectives and questions!

Kudos to the organizers Akanksha Saran Abhijat Biswas Khimya Khetarpal Reuben Aronson, Ph. D. Ruohan Zhang @neurograce Scott Neikum for putting together such an exciting program (attention-learning-workshop.github.io)!

Our first panel discussion was a great (interdisciplinary) conversation! Join us for a virtual panel on All Things Attention at 5pm CT #NeurIPS2022

Learnt a lot from speakers, authors and insightful interdisciplinary panel discussions today during our #NeurIPS2022 workshop on "bridging different perspectives of attention" AllThingsAttention. A huge thanks to everyone who participated!

You can now watch the recorded material from #NeurIPS2022 online without registration at: slideslive.com/neurips-2022