US 12,093,847 B2
Temporal processing scheme and sensorimotor information processing
Jeffrey C. Hawkins, Atherton, CA (US); Subutai Ahmad, Palo Alto, CA (US); Yuwei Cui, Lanham, MD (US); and Chetan Surpur, Cupertino, CA (US)
Assigned to Numenta, Inc., Redwood City, CA (US)
Filed by Numenta, Inc., Redwood City, CA (US)
Filed on Nov. 18, 2022, as Appl. No. 17/990,183.
Application 17/990,183 is a continuation of application No. 16/396,519, filed on Apr. 26, 2019, granted, now 11,537,922.
Application 16/396,519 is a continuation of application No. 14/662,063, filed on Mar. 18, 2015, granted, now 10,318,878, issued on Jun. 11, 2019.
Claims priority of provisional application 62/106,620, filed on Jan. 22, 2015.
Claims priority of provisional application 61/955,391, filed on Mar. 19, 2014.
Prior Publication US 2023/0111841 A1, Apr. 13, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 7/01 (2023.01); G06N 3/049 (2023.01)
CPC G06N 7/01 (2023.01) [G06N 3/049 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for temporal processing data, comprising:
detecting a plurality of spatial patterns in an input data at a first time by a first node;
generating a first sparse vector in a sparse distributed representation based on the plurality of spatial patterns detected at the first time;
predicting spatial patterns to appear in the input data at a second time subsequent to the first time responsive to (i) at least activation of more than a first predetermined number or more than a first predetermine portion of combinations of connections in the first node, and (ii) processing of the generated first sparse vector, the connections formed in the first node by different activations in the first node over time by processing a training data received prior to the input data, the connections representing relationships of temporal sequences of spatial patterns in the training data; and
generating output vectors indication output activations of the first node that vary over time based on the prediction, first output activations maintained active for a first period of time responsive to prediction associated with the first activations as being determined inaccurate, second output activations maintained active for a second period of time longer than the first period responsive to prediction associated the second output activations determined as being accurate.