US 11,706,491 B2
Deep reinforcement learning for personalized screen content optimization
Kyle Miller, Durham, NC (US)
Assigned to Rovi Guides, Inc., San Jose, CA (US)
Filed by Rovi Guides, Inc., San Jose, CA (US)
Filed on Jan. 21, 2022, as Appl. No. 17/581,490.
Application 17/581,490 is a continuation of application No. 16/893,054, filed on Jun. 4, 2020, granted, now 11,265,609.
Application 16/893,054 is a continuation of application No. 16/228,123, filed on Dec. 20, 2018, granted, now 10,715,869, issued on Jul. 14, 2020.
Prior Publication US 2022/0150589 A1, May 12, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 3/00 (2006.01); G06F 13/00 (2006.01); H04N 5/445 (2011.01); H04N 21/466 (2011.01); G06N 3/088 (2023.01)
CPC H04N 21/4666 (2013.01) [G06N 3/088 (2013.01); H04N 21/4668 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
generating for display a first set of recommended content items that were selected for a first plurality of content categories using a category selection model;
receiving a first request for a first media asset from a first content category of the plurality of content categories;
based on receiving the first request for the first media asset, increasing a reward score for the first content category;
receiving a second request for a second media asset from a second content category of the plurality of content categories;
in response to determining the second request for the second content category is identified as related to the first content category:
decreasing a reward score for the second content category;
modifying the category selection model based on the reward scores of the first content category and the second content category; and
generating for display a second set of recommended content items that were selected for a second plurality of content categories using the modified category selection model.