Devices and computer implemented methods for encoding and decoding data

1 . A computer implemented method of encoding data, the method comprising the following steps:

providing a first set of parameters that represent at least a part of the data;

determining for parameters in the first set of parameters, a weighted first sum, depending on the parameters in the first set of parameters, that is positive;

providing a first parameter that represents at least a part of the data;

determining an encoding of the data depending on a ratio between the first parameter and either the first sum or a root of a predetermined order of the first sum;

providing a second set of parameters;

determining for parameters in the second set of parameters a weighted second sum, depending on the parameters in the second set of parameters, that is positive;

providing a second parameter;

mapping the first set of parameters and/or the second set of parameters to the encoding with an encoder including weights for determining the weighted first sum and/or the weighted second sum; and

training with data that includes input data points, wherein the training includes determining with the encoder an encoding for a sequence of input data points, mapping the encoding of the sequence of input data points with a decoder to a sequence of predicted data points, wherein the encoder includes first encoder parameters for mapping the sequence of input data points to the first set of parameters and to the second set of parameters, wherein the encoder includes second encoder parameters for mapping the sequence of input data points to the first parameter and to the second parameter, and wherein the method further comprises determining the weights and/or the first encoder parameters and/or the second encoder parameters depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted at the predetermined position in the sequence of predicted data points or depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted for a different position than the predetermined position in the sequence of predicted data points.

2 . The method of encoding according to claim 1 , wherein the determining of the weighted first sum includes determining the weighted first sum of: i) even powers of the parameters, or ii) odd powers of absolute values of the parameters, or iii) even powers of the absolute values.

3 . The method of encoding according to claim 1 , wherein the method further comprises:

determining a first quantity of the encoding of the data with a first order of the parameters in the first set of parameters; and

determining a second quantity of the encoding of the data with a second order of the parameters in the first set of parameters.

4 . The method of encoding according to claim 3 , the method further comprising:

determining a third quantity of the encoding of the data depending on a ratio between the second parameter and the second sum or a root of the predetermined order of the second sum.

5 . The method according to claim 1 , wherein the data includes a plurality of input data points, wherein each input data point represents a sensor signal or a digital image, wherein the method further comprises:

receiving the input data points from or capturing the input data points by a sensor, the sensor including an image sensor, and/or a video sensor, and/or a radar sensor, and/or a LiDAR sensor, and/or an ultrasonic sensor, and/or a motion sensor, and determining a representation of the input data points, and/or

determining at least one predicted data point by decoding the representation thereof and determining a control signal depending on the at least one predicted data point for a physical system, the physical system including a computer-controlled machine, or a robot, or a vehicle, or a domestic appliance, or a power tool, or a manufacturing machine, or a personal assistant, or an access control system, or a system for conveying information, or a surveillance system, or a medical system, or a medical imaging system.

6 . A method of decoding data, the method comprising:

determining an encoding for data, the data including a sequence of input data points, the encoding including:

providing a first set of parameters that represent at least a part of the data,

determining for parameters in the first set of parameters, a weighted first sum, depending on the parameters in the first set of parameters, that is positive,

providing a first parameter that represents at least a part of the data, and

determining an encoding of the data depending on a ratio between the first parameter and either the first sum or a root of a predetermined order of the first sum;

mapping the encoding with a decoder to at least one data point, wherein the decoder includes decoder parameters for mapping the encodings of the sequence of input data points to at least one predicted data point; and

training with data that includes input data points, wherein the training includes determining with an encoder an encoding for a sequence of input data points, mapping the encoding of the sequence of input data points with the decoder to a sequence of predicted data points, wherein the decoder includes decoder parameters for mapping the encodings of the sequence of input data points to the sequence of predicted data points, and wherein the method further comprises determining the decoder parameters depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted at the predetermined position in the sequence of predicted data points or depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted for a different position than the predetermined position in the sequence of predicted data points.

7 . A device for encoding data, the device comprising:

an encoder configured to:

provide a first set of parameters that represent at least a part of the data;

determine for parameters in the first set of parameters, a weighted first sum, depending on the parameters in the first set of parameters, that is positive;

provide a first parameter that represents at least a part of the data; determine an encoding of the data depending on a ratio between the first parameter and either the first sum or a root of a predetermined order of the first sum;

provide a second set of parameters;

determine for parameters in the second set of parameters a weighted second sum, depending on the parameters in the second set of parameters, that is positive;

provide a second parameter;

map the first set of parameters and/or the second set of parameters to the encoding with the encoder including weights for determining the weighted first sum and/or the weighted second sum; and

train with data that includes input data points, wherein the training includes determining with the encoder an encoding for a sequence of input data points, mapping the encoding of the sequence of input data points with a decoder to a sequence of predicted data points, wherein the encoder includes first encoder parameters for mapping the sequence of input data points to the first set of parameters and to the second set of parameters, wherein the encoder includes second encoder parameters for mapping the sequence of input data points to the first parameter and to the second parameter, and wherein the weights and/or the first encoder parameters and/or the second encoder parameters are determined depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted at the predetermined position in the sequence of predicted data points or depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted for a different position than the predetermined position in the sequence of predicted data points.

8 . A device for decoding data, the device comprising:

an encoder configured to:

provide a first set of parameters that represent at least a part of the data, the data including a sequence of input data points;

determine for parameters in the first set of parameters, a weighted first sum, depending on the parameters in the first set of parameters, that is positive;

provide a first parameter that represents at least a part of the data; and

determine an encoding of the data depending on a ratio between the first parameter and either the first sum or a root of a predetermined order of the first sum; and

a decoder configured to map the encoding to at least one data point, wherein the decoder includes decoder parameters for mapping encodings of the sequence of input data points to at least one predicted data point, and

wherein the encoder is further configured to train with data that includes input data points, wherein the training includes determining with the encoder an encoding for a sequence of input data points, mapping the encoding of the sequence of input data points with the decoder to a sequence of predicted data points, wherein the decoder includes decoder parameters for mapping the encodings of the sequence of input data points to the sequence of predicted data points, and wherein the decoder parameters are determined depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted at the predetermined position in the sequence of predicted data points or depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted for a different position than the predetermined position in the sequence of predicted data points.

9 . A non-transitory storage medium on which is stored a computer program including machine readable instructions for encoding data, the instructions, when executed by a computer, causing the computer to perform the following steps:

providing a first set of parameters that represent at least a part of the data;

determining for parameters in the first set of parameters, a weighted first sum, depending on the parameters in the first set of parameters, that is positive;

providing a first parameter that represents at least a part of the data;

determining an encoding of the data depending on a ratio between the first parameter and either the first sum or a root of a predetermined order of the first sum;

providing a second set of parameters;

determining for parameters in the second set of parameters a weighted second sum, depending on the parameters in the second set of parameters, that is positive;

providing a second parameter;

mapping the first set of parameters and/or the second set of parameters to the encoding with an encoder including weights for determining the weighted first sum and/or the weighted second sum; and

training with data that includes input data points, wherein the training includes determining with the encoder an encoding for a sequence of input data points, mapping the encoding of the sequence of input data points with a decoder to a sequence of predicted data points, wherein the encoder includes first encoder parameters for mapping the sequence of input data points to the first set of parameters and to the second set of parameters, wherein the encoder includes second encoder parameters for mapping the sequence of input data points to the first parameter and to the second parameter, and wherein the weights and/or the first encoder parameters and/or the second encoder parameters are determined depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted at the predetermined position in the sequence of predicted data points or depending on a difference between an input data point at a predetermined position in the sequence and a data point that is predicted for a different position than the predetermined position in the sequence of predicted data points.