Thorniley, Peter

Implementations enable efficient identification of model vulnerabilities and subsequent model improvement without extensive real-world testing, leading to more robust and reliable models. Some implementations are directed to causing a training image to be processed to generate a synthetic image that includes one or more variations to the training image. The training image is one based on which a model has been trained. The variations in the synthetic image can include various changes, such as adding a new object, changing lighting, modifying the background, or altering the position of an object. The synthetic image is processed, using the model, to generate output. This output is then processed to determine whether the synthetic image is an outlier for the model. In response to determining that the synthetic image is an outlier for the model, remediating action(s) are performed, such as causing the model to be further trained based on the synthetic image.

Zusammenfassung wird geladen …

A method includes identifying, by a processor, a media stream comprising a sequence of encoded video frames and estimating, in the sequence of video frames, a share of video frames that are fragmented into respective pluralities of network packets, wherein each plurality of network packets comprises a network packet having a size below a predefined packet threshold size. Whether the share exceeds a predefined threshold value is determined and responsive to determining that the share exceeds a predefined threshold value, a target bitrate of the video stream is reduced.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a diffusion neural network using a region-aware fine-tuning process. After training, the diffusion neural network can be used to generate an image conditioned on a conditioning input.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a task. In one aspect, a method comprises: receiving a query for a task to be performed; receiving a plurality of context content items for the task; for each content item of the plurality of content items, processing an input comprising a representation of the content item using a trained compression model to generate a compressed representation of the content item comprising one or more vectors of a fixed size; generating, using the compressed representations, an aggregated compressed representation comprising one or more vectors that represents the plurality of content items; and processing an input comprising (i) the query and (ii) the aggregated compressed representation using a generative neural network to generate a response to the query.

Methods, systems, and apparatuses, including computer programs encoded on computer storage media, for training a target generative neural network over a plurality of training iterations. At each iteration, a first data item is generated by processing a conditioning input using the target generative neural network. An improvement generative neural network then processes the first data item and the conditioning input to generate a second, preferred data item. A training example is generated that includes the first and second data items and indicates that the second data item is preferred over the first. The target generative neural network is then trained on this training example. By using this iterative process to dynamically generate preference data, the described techniques improve the performance of the generative neural network beyond the limitations of static, offline datasets without requiring computationally expensive reward models or external human annotation.

Implementations are provided for an interactive machine learning methodology that allows non-expert users to use natural language to teach new skills, particularly to robots, through language grounding and understanding. In various implementations, a plurality of natural language summaries may be retrieved. Each of the natural language summaries may describe details of robotic performance of a task, and may include, or be usable to retrieve, a corresponding set of reference modulation values. A set of modulation values corresponding to a natural language request may be generated based on the plurality of natural language summaries. The natural language request may specify one or more constraints on robotic performance of the task. A robot control signal may be generated based on the generated set of modulation values.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …