[RAISE at UCR] Colloquium on AI and networking

[Vassilis Tsotras]

Colleagues,
please find below information about a Colloquium on AI and networking 
from the Department of Computer Science.
It is on Friday May 9th at 11am in the Student Success Center (SSC) Room 
329.

V. Tsotras


----------------------
Title: Explainability for Trustworthy Learning-Enabled Networked Systems

Abstract: Machine learning (ML) is revolutionizing the systems
landscape, fundamentally transforming how we design, manage, and
optimize networked systems and applications. Learning-based solutions
now outperform manually designed ones across a wide range of
environments. However, system practitioners, from cloud service
architects to network operators, often hesitate to adopt
learning-based solutions in production environments since they present
unique challenges throughout the system life cycle. For instance,
during the design phase, it is important to ensure that the operator's
design objectives are effectively captured in the trained model.
During deployment in production environments, online safety assurance
is critical. Today, it is difficult to ascertain whether the models
closely track the operator's requirements across these stages.

In this talk, I argue that explainability is key to bridging the trust
gap and enabling the broader adoption of ML in systems. Toward this
goal, I will present two explainability solutions tailored to
networked systems. First, I will introduce CrystalBox, a framework
that explains a controller's behavior in terms of the future impact on
key network performance metrics. I will demonstrate its utility in two
practical use cases: network observability and guided reward design.
Second, I will present Agua, a framework that explains a model's
decisions using high-level, human-understandable concepts (e.g.,
“volatile network conditions”). I will showcase several practical use
cases of Agua in networking environments, including debugging
unintended behaviors, identifying distribution shifts, devising
concept-based strategies for efficient retraining, and augmenting
environment-specific datasets. Together, these tools lay the
foundation for intelligent networked systems where operators can
intuitively design, debug, and adapt learning-based components through
seamless natural language interaction.

Bio: Sangeetha Abdu Jyothi is an Assistant Professor at the University
of California, Irvine. Her research interests lie at the intersection
of networking, systems, and machine learning. She received her Ph.D.
in Computer Science from the University of Illinois, Urbana-Champaign
in 2019. Her work has been recognized with several awards, including
the ACM CoNEXT Best Paper Award (2024) and the IETF/IRTF Applied
Networking Research Prize (2022). In 2022, she was named an N2Women
Rising Star in Networking and Communications.