Abstract: A computer-implemented method for training a neural network. The neural network is configured to accept an input characterizing at least one sensor measurement and provide an output characterizing a classification and/or regressions result of the at least one sensor measurement and/or a probability of the sensor measurement to occur among a set of sensor measurements or wherein the neural network is configured to provide an output characterizing a prediction of a sensor measurement for creating a training and/or test dataset for training another machine learning system. The method includes: performing a singular value decomposition of a weight matrix or a weight tensor of the neural network; grouping singular values and corresponding singular vectors into a plurality of groups based on the magnitudes of the singular values; training the neural network by adapting each group according to a distinct and non-zero learning rate and/or distinct optimizers.

Abstract: A computer-implemented method for training a neural network. The method for training includes: performing a singular value decomposition of a weight matrix or a weight tensor of the neural network; grouping singular values and corresponding singular vectors into a plurality of groups based on the magnitudes of the singular values; training the neural network by adapting the singular values and/or singular vectors of only the group corresponding to the lowest singular values or adapting the singular values and/or singular vectors of only a predefined number of groups corresponding to the lowest singular values.

Abstract: Adapting a model to tasks. The model includes a linear layer for mapping a multidimensional input of the layer depending on weights to a multidimensional output of the layer, experts, and a router gate for adapting the model to different tasks. The method includes providing the input to the router gate; determining an output of the experts depending on an output of the router gate in response to the input; modifying the model depending on the output of the experts; mapping the input with the modified layer to the output of the model; training a first expert with a first training method; training a second expert of the experts with a second training method; maintaining the weights and the second expert unchanged in the training with the first training method; and maintaining the weights and the first expert unchanged in the training with the second training method.

Abstract: Tuning weights of a neural network of a model for processing input of the model representing information about a technical system and outputting an output of the model for operating a technical system. The model includes a linear layer for mapping a multidimensional input of the layer depending on the weights to a multidimensional output of the layer. The model is configured to determine the input of the layer depending on the input of the model, and to determine the output of the model depending on the output of the layer. A method includes providing training data include the input of the model and a ground truth for the output of the model corresponding to the input of the model in the training data, providing a set of tuning methods for tuning the weights, determining the principal components decomposition of a weight matrix including the weights.

Abstract: Configuring a model. The method includes: providing the model configured for determining an output of the model depending on an output of a layer of the model, the layer being configured to map a multidimensional input of the layer depending on trained weights to a multidimensional output of the layer; iteratively arranging the trained weights in vectors of a first matrix; determining a second matrix by removing at least one vector from the first matrix; determining an input of reduced dimensions by removing from the multidimensional input the dimension that corresponds or the dimensions that correspond to the at least one vector; configuring the model with a layer of reduced dimensions configured to map the input of reduced dimensions depending on weights from the second matrix to an output of reduced dimensions; and configuring the model for determining output of the model depending on the output of reduced dimensions.

Abstract: A device and computer-implemented method for determining pixels of a synthetic image. The method comprises providing a generator that is configured to determine an output from a first input comprising a label map and a first latent code, wherein the label map comprises a mapping of at least one class to at least one of the pixels, wherein the method comprises providing the label map and a latent code, wherein the latent code comprises input data points in a latent space, providing a first direction for moving input data points in the latent space, determining the first latent code depending on at least one input data point of the latent code that is moved in the first direction, determining the synthetic image depending on an output of the generator for the first input.

Abstract: A method for generating images from a semantic map, which assigns to each pixel of the images a semantic meaning of an object to which this pixel belongs. The semantic map is provided as a map tensor comprising channels which each indicates all the pixels of the images to be generated, to which the semantic map assigns a specific semantic meaning; a set of variable pixels of the images to be generated is provided, which are to vary from one image to the next; using values taken from a random distribution, a noise tensor with channels is generated, those values of the noise tensor which relate to the set of variable pixels being reused for each image to be generated; the channels of the map tensor are merged with the channels of the noise tensor to yield an input tensor, which is mapped by a trained generator onto an image.

Abstract: Device and method for training an artificial neural network, including providing a neural network layer for an equivariant feature mapping having a plurality of output channels, grouping channels of the output channels into a number of distinct groups, wherein the output channels of each individual distinct group are organized into an individual grid defining a spatial location of each of the output channels of the individual distinct group in the grid for the individual distinct group, providing for each of the output channels of each individual distinct group, a distinct normalization function which is defined depending on the spatial location of the output channel in the grid in that this output channel is organized and depending on tunable hyperparameters for the normalization function, determining an output of the artificial neural network depending on a result of each of the distinct normalization functions, training the hyperparameters of the artificial neural network.

Abstract: A method for training a generator for images from a semantic map that assigns each pixel of the image a semantic meaning of an object to which that pixel belongs. In the method, a mixed image is generated from an image generated by the generator and a determined actual training image, in which mixed image a first genuine subset of pixels is occupied by relevant corresponding pixel values of the image generated by the generator and the remaining genuine subset of pixels is occupied by relevant corresponding pixel values of the actual training image; and the images generated by the generator, the actual training image, and at least one mixed image, which belong to the same semantic training map, are supplied to a discriminator, which is configured to distinguish images generated by the generator from actual images of the scenery predefined by the semantic training map.

Abstract: A computer-implemented method for finetuning a neural network. The method includes: providing an input to a layer of the neural network; determining a block-diagonal matrix; determining a first matrix by multiplying the block-diagonal matrix with a weight matrix of the layer, wherein the result of the multiplication is obtained by multiplying at least a plurality of blocks of the block-diagonal matrix with a respective part of the weight matrix in parallel computing operations and combining the result to form the first matrix; determining an output of the layer by multiplying the first matrix with the input of the layer; determining an output of the neural network based on the output of the layer; adapting elements of the block-diagonal matrix based on a difference of the output of the neural network and a desired output with respect to the input datum.

Abstract: A computer-implemented method for training a first machine learning system which is configured to generate an output characterizing a label map of an image. The method includes: providing first and second inputs, the first input characterizing a binary vector characterizing respective presences or absences of classes from a plurality of classes, and the second input characterizing a randomly drawn value; determining, by a first generator, an output based on the first and second inputs, the output characterizing a first label map, wherein the first label map characterizes probabilities for the classes from the plurality of classes; determining a representation of the first label map using a global pooling operation; training the first machine learning system based on a loss function, wherein the loss function characterizes an F1 loss, wherein the F1 loss characterizes a difference between the first input and the representation of the first label map.

Abstract: A device and method for evaluating a control of a generator for determining pixels of a synthetic image. The generator determining pixels of the synthetic image from a first input comprising a label map and a first latent code. The method includes providing the label map and latent code which includes input data points in a latent space; providing the control including a set of directions for moving the latent code in the latent space, determining the first latent code depending on at least one input data point of the latent code that is moved in a first direction which is selected from the set of directions, determining a distance between at least one pair of synthetic images generated by the generator for different first inputs which comprise the label map and vary by the first direction that is selected for determining the first latent code from the latent code.

Abstract: A training method for training a generator neural network configured to generate synthesized sensor data. A fidelity destroying transformation is defined configured to transform a measured sensor data to obtain a fidelity-destroyed transformed measured sensor data. A fidelity preserving transformation is defined configured to transform a measured sensor data to obtain a fidelity-preserved transformed measured sensor data.

Abstract: A method for generating synthetic video data form a text prompt, particularly for providing video data for training and/or testing and/or verifying and/or validating a machine learning model. The method includes: providing an input text prompt descriptive for the content of the video data to be generated; decomposing the provided text prompt into at least two text sub-prompts by a large language model; generating a text embedding for each of the at least two text sub-prompts; and generating synthetic video data by a Video Diffusion Model based on the generated text embeddings.

Abstract: Adapting a pretrained model to a task. The method includes providing the pretrained model including a layer configured to map a multidimensional input depending on weights to a multidimensional output, wherein a vector includes a subset of the weights that weighs the elements of the multidimensional input for a dimension of the output of the layer; providing training data and learning at least one vector of a transformation for adapting the subset depending on the training data and the output of the model, the at least one vector having unit length, and the transformation includes an outer product of the at least one vector with the transposed at least one vector, or the at least one vector is normalized to have unit length, and the transformation includes an outer product of the normalized at least one vector with the transposed normalized at least one vector.

Abstract: A computer-implemented method for generating synthetic images using a conditional diffusion model. The method involves providing a neural conditioning, which is determined by a foundation model, as input to a ControlNet. The neural conditioning and a latent input representation are then propagated through the ControlNet, and the outputs of the ControlNet are used as additional injections for the diffusion model. The latent input representation is further propagated through the diffusion model, with the additional injections from the ControlNet being injected into corresponding layers of the diffusion model during propagation.

Abstract: A technique for extracting features of an environment from image data is provided. A computer implemented method includes receiving data indicative of a visual domain of an environment; and generating a visual domain textual prompt based on the received data indicative of the visual domain of the environment. The method further includes receiving image data representative of the environment. The method further includes extracting, in particular local, features of the environment from the received image data. The extracting of the, in particular local, features is performed by a conditional feature extracting model. The extracting of the, in particular local, features is conditioned by the generated visual domain textual prompt.

Abstract: A technique for generating synthetic image data, which are usable for training, validating, and/or testing a downstream AI, in particular a downstream neural network, NN, for a body detection-related task based on sensor data. A method includes receiving visual information in relation to a body, wherein the visual information comprises a two-dimensional, 2D, skeleton representation of the body, a 2D projected (in particular dense) semantic encoding of the body, and a 2D depth map of the body. The method further includes receiving a textual prompt relating to at least one of an appearance of the body and/or environmental information relative to the body. The method further includes generating synthetic image data of the body based on the received textual prompt conditioned by the received visual information. The generating is performed by a conditional image synthesis model.

Abstract: A method for training a student neural network to adopt the behavior of a teacher neural network that is trained to perform a given processing on input images. The method includes: providing a set of training images; producing, from training image(s), one or more style-augmented versions that have the same semantic content as the original training image but differ from the original training image in their style; processing the training images and the augmented versions by the teacher neural network, and by the student neural network; evaluating, using a predetermined loss function, to which extent outputs and/or intermediate work products produced by the student neural network from each image are in agreement with the outputs and/or intermediate work products produced by the teacher neural network from the same image; and optimizing parameters that characterize the behavior of the student neural network.

Abstract: A generative adversarial network. The generative adversarial network includes: a generator configured for generating an image and a corresponding label map; a discriminator configured for determining a classification of a provided image and a provided label map, wherein the classification characterizes whether the provided image and the provided label map have been generated by the generator or not and determining the classification comprises the steps of: determining a first feature map of the provided image; masking the first feature map according to the provided label map thereby determining a masked feature map; globally pooling the masked feature map thereby determining a feature representation of the provided image masked by the provided label map; determining a classification of the image based on the feature representation.