US 12,229,007 B2
Image models to predict memory failures in computing systems
Gufeng Zhang, San Jose, CA (US); Milad Olia Hashemi, San Francisco, CA (US); and Ashish V. Naik, Los Altos, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Nov. 13, 2023, as Appl. No. 18/507,519.
Application 18/507,519 is a continuation of application No. 17/727,454, filed on Apr. 22, 2022, granted, now 11,853,161.
Prior Publication US 2024/0086281 A1, Mar. 14, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 11/10 (2006.01); G06F 11/07 (2006.01); G06F 18/214 (2023.01); G06N 3/04 (2023.01)
CPC G06F 11/1068 (2013.01) [G06F 11/0757 (2013.01); G06F 11/076 (2013.01); G06F 11/0772 (2013.01); G06F 18/214 (2023.01); G06N 3/04 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented comprising:
receiving one or more image representations generated from a log of correctable memory error data that describes correctable errors that occurred in a computer memory, wherein the one or more image representations are generated by converting the correctable memory error data to matrix codes or graphs that provide visualizations of patterns of errors that occurred in the computer memory;
processing, by an image recognition machine learning model that is trained to predict a likelihood of a future failure of a computer memory from input image representations generated by converting the log of correctable memory error data, the processing comprising:
processing, by a first convolutional neural network included in the image recognition machine learning model, a correctable error address image to obtain a first convolutional neural network output;
processing, by a second convolutional neural network included in the image recognition machine learning model, a parity syndrome image to obtain a second convolutional neural network output;
concatenating the first convolutional neural network output and the second convolutional neural network output to obtain a combined data input; and
processing, by a feed forward neural network included in the image recognition machine learning model, the combined data input to obtain an output representing the likelihood of a future failure.