An audio sample including speech and ambient sounds is transmitted to a vehicle computer. Recorded audio is received from the vehicle computer, the recorded audio including the audio sample broadcast by the vehicle computer and recorded by the vehicle computer and recognized speech from the recorded audio. The recognized speech and text of the speech are input to a machine learning program that outputs whether the recognized speech matches the text. When the output from the machine learning program indicates that the recognized speech does not match the text, the recognized speech and the text are included in a training dataset for the machine learning program.

Methods and systems are provided for detecting and responding to dangerous road users. In some aspects, a process can include steps for receiving sensor data of a detected object from an autonomous vehicle, determining whether the detected object is exhibiting a dangerous attribute, generating output data based on the determining of whether the detected object is exhibiting the dangerous attribute, and updating a machine learning model based on the output data relating to the dangerous attribute.

System and methods enable vehicles to make ethical/empathetic driving decisions by using deep learning aided location intelligence. The systems and methods identify moral islands/complex driving scenarios where a complex ethical decision is required. A Generative Adversarial Network (GAN) is used to generate synthetic training data to capture varied ethically complex driving situations. Embodiments train a deep learning model (ETHNET) that is configured to output one or more driving decisions to be taken when a vehicle comes across an ethically complex driving situations in the real world.

Embodiments including a method and apparatus for correction of a global navigation satellite system (GNSS) are described. In one example, the apparatus includes a communication interface and a processor. The communication interface is configured to a plurality of GNSS signals. The GNSS signals may include at least one almanac value and at least one ephemeris value. The processor is configured to generate a spatio-temporal graph model based on the at least one almanac value, the at least one ephemeris value, and a predetermined offset value for a base location. The spatio-temporal graph model analyzes subsequent GNSS signals to determined a predicted offset or a corrected GNSS position.

The subject matter described herein includes methods, systems, and computer readable media for mask embedding for realistic high-resolution image synthesis. According to one method for mask embedding for realistic high-resolution image synthesis includes receiving, as input, a mask embedding vector and a latent features vector, wherein the mask embedding vector acts as a semantic constraint; generating, using a trained image synthesis algorithm and the input, a realistic image, wherein the realistic image is constrained by the mask embedding vector; and outputting, by the trained image synthesis algorithm, the realistic image to a display or a storage device.

A medical image segmentation method based on a U-Net, including: sending real segmentation image and original image to a generative adversarial network for data enhancement to generate a composite image with a label; then putting the composite image into original data set to obtain an expanded data set, and sending the expanded data set to improved multi-feature fusion segmentation network for training. A Dilated Convolution Module is added between the shallow and deep feature skip connections of the segmentation network to obtain receptive fields with different sizes, which enhances the fusion of detail information and deep semantics, improves the adaptability to the size of the segmentation target, and improves the medical image segmentation accuracy. The over-fitting problem that occurs when training the segmentation network is alleviated by using the expanded data set of the generative adversarial network.

A computer system (which may include one or more computers) that chooses or selects one or more criteria for when to terminate training of a neural network is described. During operation, the computer system may choose or select the one or more criteria for when to terminate the training of the neural network, where the one or more criteria are based at least in part on a measure corresponding to a local geometry of a loss landscape at or proximate to a current location in the loss landscape. Note that one of the one or more criteria may include: a trace of a Hessian matrix associated with a loss function dropping below a threshold, or a ratio between an operator norm of the Hessian matrix and a curvature of the loss function at the current location in the loss landscape reaching a second threshold.

Systems and methods for architecture embeddings for efficient dynamic synthetic data generation are disclosed. The disclosed systems and methods may include a system for generating synthetic data configured to perform operations. The operations may include retrieving a set of rules associated with a first data profile and generating, by executing a hyperparameter search, a plurality of hyperparameter sets for generative adversarial networks (GANs) that satisfy the set of rules. The operations may include generating mappings between the hyperparameter sets and the first data profile and storing the mappings in a hyperparameter library. The operations may include receiving a request for synthetic data, the request indicating a second data profile and selecting, from the mappings in the hyperparameter library, a hyperparameter set mapped to the second data profile. The operations may include building a GAN using the selected hyperparameter set and generating, using the GAN, a synthetic data set.

A method includes obtaining (such as accessing, receiving, acquiring, etc.), using at least one processor of an electronic device, a machine learning model trained to process input data and generate output data over at least one range of values associated with one or more control variables. The method also includes providing, using the at least one processor, specified input data to the machine learning model and providing, using the at least one processor, one or more specified values of the one or more control variables to the machine learning model. The one or more specified values of the one or more control variables are within the at least one range of values. The method further includes performing inferencing using the machine learning model to process the specified input data and generate specified output data. The inferencing is controlled based on the one or more specified values of the control variable(s).

Machine learning adversarial campaign mitigation on a computing device. The method may include deploying an original machine learning model in a model environment associated with a client device; deploying a classification monitor in the model environment to monitor classification decision outputs in the machine learning model; detecting, by the classification monitor, a campaign of adversarial classification decision outputs in the machine learning model; applying a transformation function to the machine learning model in the model environment to transform the adversarial classification decision outputs to thwart the campaign of adversarial classification decision outputs; determining a malicious attack on the client device based in part on detecting the campaign of adversarial classification decision outputs; and implementing a security action to protect the computing device against the malicious attack.