公开(公告)号：US20240119253A1

公开(公告)日：2024-04-11

申请号：US17957891

申请日：2022-09-30

Applicant: SAP SE

Inventor： Florian Knoerzer ,  Swen Koenig ,  Dominic Hehn ,  Mustafa Aktan ,  Jocelyn Borella ,  Naseer Muhammad

IPC: G06N3/04

CPC classification number: G06N3/04

Abstract: In an example embodiment, an additional classifier is introduced to an autoencoder neural network. The additional classifier performs an additional classification task during the training and testing phases of the autoencoder neural network. More precisely, the autoencoder neural network learns to classify the domain (or origin) of each specific input sample. This leads to additional contextual awareness in the autoencoder neural network, which improves the reconstruction quality during both the training and testing phases. Thus, the technical problem of decreased autoencoder neural network reconstruction quality caused by high data variance is addressed.

公开(公告)号：US12105847B2

公开(公告)日：2024-10-01

申请号：US17550634

申请日：2021-12-14

Applicant: SAP SE

Inventor： Benjamin Weggenmann ,  Martin Haerterich ,  Florian Knoerzer

IPC: G06F21/62

CPC classification number: G06F21/6254

Abstract: Techniques for implementing a differentially private variational autoencoder for data obfuscation are disclosed. In some embodiments, a computer system performs operations comprising: encoding input data into a latent space representation of the input data, the encoding of the input data comprising: inferring latent space parameters of a latent space distribution based on the input data, the latent space parameters comprising a mean and a standard deviation, the inferring of the latent space parameters comprising bounding the mean within a finite space and using a global value for the standard deviation, the global value being independent of the input data; and sampling data from the latent space distribution; and decoding the sampled data of the latent space representation into output data.

公开(公告)号：US20220070150A1

公开(公告)日：2022-03-03

申请号：US17010501

申请日：2020-09-02

Applicant: SAP SE

Inventor： Martin Haerterich ,  Benjamin Weggenmann ,  Florian Knoerzer

IPC: H04L29/06 ,  G06F21/62 ,  G06N3/04 ,  G06N3/08

Abstract: Various examples are directed to systems and methods for obscuring personal information in a sensor data stream. A system may apply an encoder model to the sensor data stream to generate a latent space representation of the sensor data stream. The system may also apply a noise-scaling parameter to the latent space representation of the sensor data stream and apply a decoder model to the latent space representation of the sensor data stream to generate an obscured data stream.

公开(公告)号：US20240427936A1

公开(公告)日：2024-12-26

申请号：US18827444

申请日：2024-09-06

Applicant: SAP SE

Inventor： Banjamin Weggenmann ,  Martin Haerterich ,  Florian Knoerzer

IPC: G06F21/62

Abstract: Techniques for implementing a differentially private variational autoencoder for data obfuscation are disclosed. In some embodiments, a computer system performs operations comprising: encoding input data into a latent space representation of the input data, the encoding of the input data comprising: inferring latent space parameters of a latent space distribution based on the input data, the latent space parameters comprising a mean and a standard deviation, the inferring of the latent space parameters comprising bounding the mean within a finite space and using a global value for the standard deviation, the global value being independent of the input data; and sampling data from the latent space distribution; and decoding the sampled data of the latent space representation into output data.

公开(公告)号：US20240354370A1

公开(公告)日：2024-10-24

申请号：US18305591

申请日：2023-04-24

Applicant: SAP SE

Inventor： Dominic Hehn ,  Florian Knoerzer ,  Swen Koenig ,  Jocelyn Borella ,  Mustafa Aktan ,  Naseer Muhammad

IPC: G06F17/18

CPC classification number: G06F17/18

Abstract: Embodiments determine change points within received time series data exhibiting a natural trend. A first candidate change point comprising an earlier time and a first value, and a second candidate change point comprising a later time and a second value, are received with the time series data. A rule is executed upon the first candidate change point to calculate a first score, and executed upon the second candidate change point to calculate a second score. The rule comprises a primary criterion for a change direction relative to the natural trend, a secondary criterion for a change position within the time series data, and a tertiary criterion for a change magnitude. The first score is compared to the second score to select the first candidate change point or the second candidate change point as a determined change point. The determined change point is stored for use in subsequent data analysis.

公开(公告)号：US20230185962A1

公开(公告)日：2023-06-15

申请号：US17550634

申请日：2021-12-14

Applicant: SAP SE

Inventor： Benjamin Weggenmann ,  Martin Haerterich ,  Florian Knoerzer

IPC: G06F21/62

CPC classification number: G06F21/6254

Abstract: Techniques for implementing a differentially private variational autoencoder for data obfuscation are disclosed. In some embodiments, a computer system performs operations comprising: encoding input data into a latent space representation of the input data, the encoding of the input data comprising: inferring latent space parameters of a latent space distribution based on the input data, the latent space parameters comprising a mean and a standard deviation, the inferring of the latent space parameters comprising bounding the mean within a finite space and using a global value for the standard deviation, the global value being independent of the input data; and sampling data from the latent space distribution; and decoding the sampled data of the latent space representation into output data.