US 12,243,072 B2
Segment modeling for machine learning using tensor train decompositions
Ajith Muralidharan, Sunnyvale, CA (US); Ankan Saha, San Francisco, CA (US); and Prakruthi Prabhakar, Foster City, CA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jan. 6, 2023, as Appl. No. 18/094,027.
Prior Publication US 2024/0232939 A1, Jul. 11, 2024
Int. Cl. G06Q 30/02 (2023.01); G06N 3/08 (2023.01); G06Q 30/0242 (2023.01)
CPC G06Q 30/0246 (2013.01) [G06N 3/08 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A system comprising:
a computer-readable medium having instructions stored thereon, which, when executed by a processor, cause the system to perform operations comprising:
accessing training data comprising information about one or more users of an online network and one or more content items of the online network;
obtaining one or more segments of the training data, wherein the one or more segments are each a set of one or more common characteristics of the training data; and
training a neural network by feeding the training data into the neural network, the neural network containing a plurality of fully-connected layers, one of the fully-connected layers being a tensor train layer designed to convert a matrix associated with the at least one of the fully connected layers into a tensor using tensor train decomposition, each of a plurality of dimensions of the tensor corresponding to a respective property used to define a different segment of the training data, the tensor train layer comprising a last layer in the neural network that receives a side input comprising segment information for the one or more segments.