[Coral_current] Recruiting students to meet tomorrow's faculty candidate, 5-5:30 in ITE325b (fwd)

[Tim Oates]

If you can meet with the faculty candidate tomorrow it would be very
helpful.  See below.  He seems like a great AI person that we'd like to
have at UMBC.

  - tim


---------------------------------------
Tim Oates, Professor
Department of CS and EE
University of Maryland Baltimore County
(410) 455-3082
https://coral-lab.umbc.edu/oates/

---------- Forwarded message ----------
Date: Tue, 14 Mar 2023 13:05:58 -0400
From: Tim Finin <finin at cs.umbc.edu>
To: Roberto Yus <ryus at umbc.edu>, Manas Gaur <manas at umbc.edu>,
     Frank Ferraro <ferraro at umbc.edu>, Adam Bargteil <adamb at umbc.edu>,
     Don Engel <donengel at umbc.edu>, James Oates <oates at cs.umbc.edu>
Cc: Nilanjan Banerjee <nilanb at umbc.edu>
Subject: Recruiting students to meet tomorrow's faculty candidate,
     5-5:30 in ITE325b

Thanks for signing up to meet with faculty candidate Tejas Gokhale tomorrow
(Wed., 3/15). He expressed a strong interest in meeting with some of our
students, and we've scheduled 5-5:30 for him to meet with students in our
meeting room, ITE 325b. Can you share this opportunity with any students you
have or know who might be interested and encourage them to join the group
meeting with him? His talk announcement is below in case you want to send it
to your students.

____________________________________________________________________________

Towards Reliable Semantic Vision

Tejas Gokhale
Arizona State University

Talk: 12-1 Wednesday, 15 March 2023, ITE 459 Meet with students: 5-5:30pm in
ITE 325


Models that learn from data are widely and rapidly being deployed today for
real-world use, but they suffer from unforeseen failures that limit their
reliability. In this talk, I will describe my work toward building reliable
computer vision systems that can leverage domain knowledge. First, I will
show how data transformations can be discovered during training to improve
the robustness of image classifiers to distribution shift. Second, I will
demonstrate intriguing failure modes in the multimodal (image + text)
setting and how training paradigms can mitigate these failures guided by
domain knowledge of logic and linguistics. Finally, I will discuss
strategies for enhancing semantic vision systems with complex visual
reasoning abilities, which will be crucial for reliability in the imminent
future of continual and collaborative multimodal learning agents.


Tejas Gokhale is a Ph.D. candidate at Arizona State University, co-advised
by Yezhou Yang and Chitta Baral. He received his M.S. from Carnegie Mellon
University in 2017. His research is broadly in the areas of computer vision,
natural language processing, and machine learning, with a focus on
robustness and reliability. He is particularly interested in semantic vision
systems that can learn from human language and reason about the visual
world. Tejas' research, mentorship, and scientific service have been
recognized by the ASU Engineering Graduate Fellowship, SCAI Doctoral
Fellowship, GPSA Outstanding Mentor Award, top reviewer awards at ICLR and
NeurIPS. He is the lead organizer of the O-DRUM workshop at CVPR.