{"id":3855,"date":"2025-09-15T10:10:40","date_gmt":"2025-09-15T15:10:40","guid":{"rendered":"https:\/\/wordpress.cels.anl.gov\/lans-seminars\/?post_type=tribe_events&p=3855"},"modified":"2025-09-15T10:10:40","modified_gmt":"2025-09-15T15:10:40","slug":"lans-seminar-182","status":"publish","type":"tribe_events","link":"https:\/\/wordpress.cels.anl.gov\/lans-seminars\/event\/lans-seminar-182\/","title":{"rendered":"LANS Seminar"},"content":{"rendered":"
\n

Seminar Title<\/strong>: Tucker tensor train Taylor series approximation of high dimensional implicit mappings<\/p>\n

Speaker: <\/strong>Nick Alger,Researcher, Oden Institute, UT Austin Working on Numerical Methods for Solving Inverse Problems Governed by PDEs<\/p>\n

Date:<\/strong> Thursday, September 25, 2025<\/p>\n

Time:<\/strong> 2:30 PM-3:30 PM (In-Person)<\/p>\n

Location:<\/strong>\u00a0Hybrid, Bldg. 240, Conference Room 4301<\/p>\n

Host:<\/strong> Vishwas Rao<\/p>\n

Description<\/strong>: We present an efficient method for constructing high order Taylor <\/span>series surrogate models for high dimensional mappings that depend <\/span>implicitly on the solution of a system of nonlinear equations, e.g., a <\/span>partial differential equation. High order Taylor series are <\/span>traditionally considered intractable here because the derivative <\/span>tensors are extremely large, and are only accessible through <\/span>multilinear actions on vectors. We overcome these challenges using a <\/span>“Tucker tensor train Taylor series” surrogate model, in which linear <\/span>(Tucker) dimension reduction is performed on the input and output <\/span>spaces, and the derivative tensors are approximated by tensor trains <\/span>in the reduced subspaces. The Tucker bases are constructed using <\/span>randomized sketching, and the tensor trains are fit to directionally <\/span>symmetric action data using a Riemannian manifold Newton method. We <\/span>justify the model theoretically, and provide numerical evidence for <\/span>the effectiveness of the proposed method.<\/span><\/p>\n

Bio: <\/strong>Nick Alger<\/span> is a researcher in the Oden Institute at UT Austin working <\/span>on numerical methods for solving inverse problems governed by PDEs. He <\/span>received his Ph.D. and M.S. in computational science from UT Austin, <\/span>and B.S. in physics from Harvey Mudd College. <\/span><\/p>\n

Please note that the meeting URL for this event can be seen on the cels-seminars website which requires an Argonne login.<\/em><\/p>\n

\n

See all upcoming talks at\u00a0https:\/\/www.anl.gov\/mcs\/lans-seminars<\/a><\/p>\n<\/div>\n<\/div>\n

\n