For those who are interested in joining remotely, here is the Zoom link:
https://cispa-de.zoom-x.de/my/muandet
Best, Krikamol
---
Krikamol Muandet, Dr. rer. nat. | Tenure-Track Faculty CISPA — Helmholtz Center for Information Security Stuhlsatzenhaus 5, 66123 Saarbrücken, Germany +49 681 87083 2558 | muandet@cispa.de | https://krikamol.org
On 3. Dec 2024, at 08:25, Krikamol Muandet muandet@cispa.de wrote:
Dear all,
A final reminder that this guest talk will take place TODAY at 10:30am. Looking forward to seeing you there.
Best, Krikamol
On 2. Dec 2024, at 16:28, Krikamol Muandet muandet@cispa.de wrote:
Dear all,
A gentle reminder that this guest talk will take place tomorrow.
Best, Krikamol
On 30. Nov 2024, at 07:33, Krikamol Muandet muandet@cispa.de wrote:
Dear all,
Amin Charusaie (https://charusaie.github.io/) from Max Planck Institute for Intelligent Systems (MPI-IS) will visit us during Dec 2-3 and will talk about his research. You are cordially invited to attend the talk. Please feel free to forward this information to anyone who might be interested.
Title: Optimal Multi-Objective Learn-to-Defer: Possibility, Complexity, and a Post-Processing Framework Location: Seminar room 1.01, CISPA D2 Date/Time: Tuesday, December 3rd at 10:30am
Abstract:
Learn-to-Defer (L2D) is a paradigm that enables learning algorithms to work not in isolation but as a team with human experts. In this paradigm, we permit the system to defer a subset of its tasks to the expert. Although there are currently systems that follow this paradigm and are designed to optimize the accuracy of the final human-AI team, the general methodology for developing such systems under a set of constraints (e.g., algorithmic fairness, expert intervention budget, defer of anomaly, etc.) remains largely unexplored. In this presentation, I discuss the complexity of obtaining an optimal deterministic solution to multi-objective L2D, possibility of learning deferral labels, and achieve the optimal random solution via a d-dimensional generalization to the fundamental lemma of Neyman and Pearson. I further discuss the implications of such generalization to a variety of multi-objective learning problems beyond L2D. Finally, I provide experimental results that show the effectiveness of the introduced method on a series of L2D datasets.
If you would like to meet him either in person or virtually, please contact me via muandet@cispa.de.
Best, Krikamol
Krikamol Muandet, Dr. rer. nat. | Tenure-Track Faculty CISPA — Helmholtz Center for Information Security Stuhlsatzenhaus 5, 66123 Saarbrücken, Germany +49 681 87083 2558 | muandet@cispa.de | https://krikamol.org
Ml-reading-group mailing list -- ml-reading-group@lists.saarland-informatics-campus.de To unsubscribe send an email to ml-reading-group-leave@lists.saarland-informatics-campus.de