BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:129@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20250623T160000
DTEND;TZID=Asia/Kolkata:20250623T170000
DTSTAMP:20250619T120906Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-june-23rd-0400-anchoring
 -deep-models-for-reliable-and-robust-machine-learning/
SUMMARY:{Seminar} @ CDS: #102\, June 23rd\, 04:00: "Anchoring Deep Models f
 or Reliable and Robust Machine Learning."
DESCRIPTION:Department of Computational and Data Sciences\nDepartment Semin
 ar\n==================================================================\n\n
 Speaker : Dr. Vivek Sivaraman Narayanaswamy\nTitle   : "Anchoring Deep M
 odels for Reliable and Robust Machine Learning"\nDate &amp\; Time: June 23
 rd\, 2025 (Monday)\, 04:00 PM\nVenue : # 102\, CDS Seminar Hall\n\n\n\nABS
 TRACT\n\nAs machine learning and AI systems are increasingly deployed in h
 igh-stakes\, real-world settings\, ensuring their reliability under divers
 e conditions is more important than ever. Reliability spans multiple dimen
 sions\, including robustness to corruptions and distribution shifts\, out-
 of-distribution (OOD) detection\, calibration\, generalization\, extrapola
 tion\, uncertainty quantification and failure detection. Despite ongoing p
 rogress\, achieving consistent performance across these axes remains a fun
 damental challenge.\n\nThis talk introduces anchoring\, a simple yet effec
 tive training paradigm for deep networks. Anchoring reparameterizes each i
 nput as a tuple comprising a reference sample and a residual (the differen
 ce between the input and the reference)\, while preserving the original la
 bel. Both reference and input samples are drawn from the training distribu
 tion. Though lightweight in design\, this modification enables the model t
 o explore a richer hypothesis space\, leading to better convergence and im
 provements across multiple reliability dimensions without requiring signif
 icant architectural changes. Importantly\, anchoring is model-agnostic and
  applicable even to state-of-the-art vision transformers. The talk will pr
 esent key insights about anchoring across tasks and modalities\, demonstra
 te its consistent impact on reliability\, and discuss its integration into
  standard training pipelines along with promising future directions.\n\nBI
 O: Vivek Sivaraman Narayanaswamy is a Machine Learning Research Scientist 
 with the Machine Intelligence Group at the Center for Applied Scientific C
 omputing (CASC). His work lies at the intersection of deep learning\, comp
 uter vision\, and scientific machine learning with a focus on building tru
 stworthy AI systems. He develops neural network surrogates for complex sim
 ulations\, embedding uncertainty quantification to enhance reliability and
  designing protocols for better extrapolation\, calibration\, and anomaly 
 detection. He earned his Ph.D. in Electrical Engineering from Arizona Stat
 e University\, USA in 2023\, where his research centered on improving the 
 fidelity and reliability of deep learning models. His work has been publis
 hed at top-tier AI conferences such as NeurIPS\, ICML\, ECCV\, AAAI\, and 
 ICASSP. He is also passionate about mentorship and has co-supervised stude
 nt research on foundation models\, conformal prediction\, and scientific M
 L.\n\nHost Faculty: Yogesh Simmhan\n\n\n\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20240623T160000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR