US 12,073,305 B2
Deep multi-task representation learning
Mohamed R. Amer, Brooklyn, NY (US); Timothy J. Shields, Houston, TX (US); Amir Tamrakar, New Brunswick, NJ (US); Max Ehrlich, Princeton, NJ (US); and Timur Almaev, Nottingham (GB)
Assigned to SRI International, Menlo Park, CA (US)
Appl. No. 16/085,859
Filed by SRI International, Menlo Park, CA (US)
PCT Filed Mar. 17, 2017, PCT No. PCT/US2017/022902
§ 371(c)(1), (2) Date Sep. 17, 2018,
PCT Pub. No. WO2017/161233, PCT Pub. Date Sep. 21, 2017.
Claims priority of provisional application 62/309,804, filed on Mar. 17, 2016.
Prior Publication US 2019/0034814 A1, Jan. 31, 2019
Int. Cl. G06N 3/045 (2023.01); G06F 18/2132 (2023.01); G06F 18/24 (2023.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01)
CPC G06N 3/045 (2023.01) [G06F 18/2132 (2023.01); G06F 18/24 (2023.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01)] 12 Claims
OG exemplary drawing
 
6. A method for classifying data, the method comprising, with a computing system comprising one or more computing devices:
accessing a set of instances of data having a plurality of modalities, wherein the plurality of modalities are associated with a plurality of tasks;
algorithmically classifying the data using a dynamic hybrid model that uses joint optimization of generative and discriminative processes as a single, non-staged framework including parameters learned at the same time and trained in unison; and
learning an inference model using an iterative bottom-up reconstructive and top-down generative approach.