Using neural networks to encode log data

Patent metadata

Field	Value
Publication number	US-12499303-B2
Application number	US-202418658362-A
Country	US
Kind code	B2
Filing date	May 8, 2024
Priority date	Apr 29, 2024
Publication date	Dec 16, 2025
Grant date	Dec 16, 2025

Abstract

Methods, systems, and machine-readable mediums to perform a neural network to encode log data. In at least one embodiment, a processor comprising one or more circuits to encode at least one log message, at least in part, by encoding a first type of information in the at least one log message to obtain a first encoding, encoding a second type of information in the at least one log message to obtain a second encoding, and obtaining a resultant encoding at least in part by combing at least the first and second encodings.

First claim

What is claimed is:

1 . A method comprising:
encoding at least one log message, at least in part, by: using at least one semantic encoder to encode a first type of information in the at least one log message to obtain a first encoding; using at least one numerical encoder to encode a second type of information in the at least one log message to obtain a second encoding; using at least one categorical encoder to encode a third type of information in the at least one log message to obtain a third encoding; and performing an attention layer to assign weights to feature embeddings of the first, second, and third encodings and to generate a resultant encoding by using the weights to combine the feature embeddings.

2 . The method of claim 1 , further comprising:
providing the resultant encoding to at least one neural network trained using adversarial learning to generate a first encoding; providing the first encoding to at least one classifier to generate a second encoding; and combining the second encoding with at least one of topology information or telemetry information.

3 . The method of claim 1 , further comprising:
identifying the first, second, and third types of information in the at least one log message before encoding the first, second, and third types of information.

4 . The method of claim 1 , wherein the third type of information includes a priority associated with the at least one log message.

5 . The method of claim 1 , wherein the resultant encoding includes a vector encoding.

6 . The method of claim 1 , wherein the attention layer is to generate the weights using alignment scores generated between a query vector and the first, second, and third encoding.

7 . The method of claim 1 , further comprising:
using at least one other encoder to generate one or more encodings based, at least in part, on the resultant encoding and at least one of telemetry information or topology information, and using the one or more encodings to perform at least one of anomaly detection, incident prediction, root cause analysis, or observation generation.

8 . The method of claim 1 , further comprising:
associating the at least one log message with at least one identifier of at least one network node; and using at least one other encoder to generate one or more encodings based, at least in part, on the resultant encoding and the at least one identifier.

9 . The method of claim 1 , wherein the method is performed by at least one of:
a first system to perform neural network training operations; a second system to perform deep learning operations; a third system to generate data; a fourth system implemented at least partially in a data center; or a fifth system implemented at least partially using cloud computing resources.

10 . The method of claim 1 , further comprises:
using the resultant encoding to perform anomaly detection.

11 . A processor comprising:
one or more circuits to encode at least one log message, at least in part, by: using at least one semantic encoder to encode a first type of information in the at least one log message to obtain a first encoding; using at least one numerical encoder to encode a second type of information in the at least one log message to obtain a second encoding; using at least one categorical encoder to encode a third type of information in the at least one log message to obtain a third encoding; and performing an attention layer to assign weights to feature embeddings of the first, second, and third encodings and to generate a resultant encoding by using the weights to combine the feature embeddings.

12 . The processor of claim 11 , wherein the one or more circuits to are to identify the first, second, and third types of information in the at least one log message before encoding the first, second, and third types of information.

13 . The processor of claim 11 , wherein the one or more circuits are to:
use at least one neural network to generate one or more encodings based, at least in part, on the resultant encoding and at least one of telemetry information or topology information, and use the one or more encodings to perform at least one of anomaly detection, incident prediction, root cause analysis, or observation generation.

14 . The processor of claim 11 , wherein the one or more circuits are to use at least one neural network to implement at least one of the at least one semantic encoder, the at least one numerical encoder, or the at least one categorical encoder.

15 . The processor of claim 11 , wherein the resultant encoding includes a vector encoding.

16 . A system comprising:
one or more processors to encode at least one log message, at least in part, by: using at least one semantic encoder to encode a first type of information in the at least one log message to obtain a first encoding; using at least one numerical encoder to encode a second type of information in the at least one log message to obtain a second encoding; using at least one categorical encoder to encode a third type of information in the at least one log message to obtain a third encoding; and performing an attention layer to assign weights to feature embeddings of the first, second, and third encodings and to generate a resultant encoding by using the weights to combine the feature embeddings.

17 . The system of claim 16 , wherein the one or more processors are to:
use at least one neural network to generate one or more encodings based, at least in part, on the resultant encoding and at least one of telemetry information or topology information, and use the one or more encodings to perform at least one of anomaly detection, incident prediction, root cause analysis, or observation generation.

18 . The system of claim 16 , wherein the one or more processors are to identify the first, second, and third types of information in the at least one log message before encoding the first, second, and third types of information.

19 . The system of claim 16 , wherein the one or more processors are to use at least one neural network to implement at least one of the at least one semantic encoder, the at least one numerical encoder, or the at least one categorical encoder.

20 . The system of claim 16 , wherein the one or more processors are to use the resultant encoding to perform anomaly detection.

Assignees

• Mellanox Technologies Ltd

Inventors

• Shavit Yoli
• Zahavi Eitan
• Mataev Gary
• Shteingart Hanan
• Puget Jean-Francois
• Binshtock Zachi

Classifications

• G06F18/2433

  Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection · CPC title

• G06F16/353

  into predefined classes · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title

• G06F40/126 Primary

  Character encoding · CPC title

• H03M7/3082 Primary

  Vector coding (for television signals, see H04N19/94) · CPC title