Abstract: Methods and systems for time marking of media items at a platform using machine-learning are provided herein. A media item is provided to users of a platform. An indication of the identified media item is provided as input to a machine-learning model that is trained using different feature types of historical media items to predict a plurality of content segments of a given media item each depicting, to the one or more users, a distinct section of the media item. One or more outputs of the machine-learning model are obtained comprising time marks identifying each of the plurality of content segments of the media item. Each of the plurality of content segments are associated with a segment start indicator for a timeline of the media item. The media item and an indication of each segment start indicator is provided for presentation to at least one user.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for neural network adaptive beamforming for multichannel speech recognition are disclosed. In one aspect, a method includes the actions of receiving a first channel of audio data corresponding to an utterance and a second channel of audio data corresponding to the utterance. The actions further include generating a first set of filter parameters for a first filter based on the first channel of audio data and the second channel of audio data and a second set of filter parameters for a second filter based on the first channel of audio data and the second channel of audio data. The actions further include generating a single combined channel of audio data. The actions further include inputting the audio data to a neural network. The actions further include providing a transcription for the utterance.

Abstract: Example embodiments relate to techniques for training artificial neural networks or oilier machine-learning encoders to accurately predict the pitch of input audio samples in a semitone or otherwise logarithmically-scaled pitch space. An example method may include generating, from a sample of audio data, two training samples by applying two different pitch shifts to the sample of audio training data. This can be done by converting the sample of audio data into the frequency domain and then shifting the transformed data. These known shifts are then compared to the predicted pitches generated by applying the two training samples to the encoder. The encoder is then updated based on the comparison, such that the relative pitch output by the encoder is improved with respect to accuracy. One or more audio samples, labeled with absolute pitch values, can then be used to calibrate the relative pitch values generated by the trained encoder.

Abstract: Methods, systems, and media for presenting content based on a generic rating are provided.

Abstract: A method for data storage includes assigning multiple different portions of a given object, which is stored in non-volatile storage, for access by multiple respective software modules running on one or more processors. Storage operations are performed by the software modules in the portions of the given object. Respective local values, which are indicative of most recent times the storage operations were performed in the given object by the respective software modules, are updated by the software modules. A global value is updated for the given object, by consolidating updates of the local values performed by the software modules.

Abstract: The methods and systems described herein provide for depth-aware image editing and interactive features. In particular, a computer application may provide image-related features that utilize a combination of a (a) the depth map, and (b) segmentation data to process one or more images, and generate an edited version of the one or more images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for overlaying content on video streams. In one aspect, a category of a video is identified. Confidence scores are determined that each indicate a likelihood that a location in a frame includes a feature of a feature type. A weight for each feature type is determined, based on the category, that reflects an importance of not occluding features of the feature type. Confidence scores are adjusted, for each feature type, based on the weight for the feature type, to generate adjusted confidence scores. The adjusted confidence scores are aggregated for each location for each frame to generate aggregated and adjusted confidence scores. A location at which to position overlaid content during video display is determined based on the aggregated and adjusted confidence scores. The overlaid content is provided for display at the determined location in the video.

Abstract: This technology relates to transferring state information between processes or active software programs in a computing environment where a new instance of a process or software program may receive such state information even after an original or old instance of the process or software program that owned the state information has terminated either naturally or unnaturally.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.

Abstract: A method for generating random numbers includes initializing a pseudo-random number generator (PRNG) having a state of 2048 bits comprising inner bits and outer bits, the inner bits comprising the first 128 bits of the 2048 bits and the outer bits comprising the remaining bits of the 2048 bits. The method also includes retrieving AES round keys from a key source, and for a threshold number of times, executing a round function using the AES round keys by XOR'ing odd-numbered branches of a Feistel network having 16 branches of 128 bits with a function of corresponding even-numbered neighbor branches of the Feistel network, and shuffling each branch of 128 bits into a prescribed order. The method also includes executing an XOR of the inner bits of the permuted state with the inner bits of a previous state.

Abstract: A first segment of a probe media item that is transmitted as a first live-stream of an event is received. After a first delay period, whether the first segment of the probe media item is similar to a first segment of a first reference media item that is transmitted as a second live-stream of the event and received subsequent to the probe media item is determined. Responsive to determining, after the first delay period, that the first segment is similar to the first segment of the first reference media item, a remedial action in association with the probe media item is performed.

Abstract: The subject matter of this specification can be implemented in, among other things, a computer-implemented method that includes identifying metadata related to television programming being presented on a display device. The method further includes extracting one or more keywords from the metadata. The method further includes generating multiple search suggestions based on the keywords and first search results based on one or more of the search suggestions. The method further includes presenting the search suggestions and the first search results together on the display device.

Abstract: Various devices, systems and methods for performing targeted sleep tracking are presented herein. Multiple digital radar data streams from multiple antennas may be received. A direction optimization process may be performed to determine a first weighting and a second weighting. The first weighting may be applied to a first digital radar data stream. The second weighting may be applied to a second digital radar data stream. The weighted first digital radar data stream and the weighted second digital radar data stream may be combined to create a first directionally-targeted radar data stream. A sleep analysis can be performed based on the first directionally-targeted radar data stream. Sleep data may be output for a user based on the performed sleep analysis.

Abstract: A method is provided that includes controlling a robotic gripping device to cause a plurality of digits of the robotic gripping device to move towards each other in an attempt to grasp an object. The method also includes receiving, from at least one non-contact sensor on the robotic gripping device, first sensor data indicative of a region between the plurality of digits of the robotic gripping device. The method further includes receiving, from the at least one non-contact sensor on the robotic gripping device, second sensor data indicative of the region between the plurality of digits of the robotic gripping device, where the second sensor data is based on a different sensing modality than the first sensor data. The method additionally includes determining, using an object-in-hand classifier that takes as input the first sensor data and the second sensor data, a result of the attempt to grasp the object.

Abstract: A control device of voice distribution including: at least one voice processing module arranged to—receive as input an audio signal including a first vocal message, and—provide as output an audio signal including a second vocal message, the first and second vocal messages being different one from the other and the second vocal message resulting from a processing of the first vocal message; a communication module arranged to establish and simultaneously manage a wireless, bidirectional and audio link with each one of a plurality of auxiliary devices, each link being connected to the input and/or the output of at least one voice processing module.

Abstract: Deep recurrent neural networks applied to speech recognition. The deep recurrent neural networks (RNNs) are preferably implemented by stacked long short-term memory bidirectional RNNs. The RNNs are trained using end-to-end training with suitable regularisation.

Abstract: A method for determining synthetic speech includes receiving audio data characterizing speech in audio data obtained by a user device. The method also includes generating, using a trained self-supervised model, a plurality of audio features vectors each representative of audio features of a portion of the audio data. The method also includes generating, using a shallow discriminator model, a score indicating a presence of synthetic speech in the audio data based on the corresponding audio features of each audio feature vector of the plurality of audio feature vectors. The method also includes determining whether the score satisfies a synthetic speech detection threshold. When the score satisfies the synthetic speech detection threshold, the method includes determining that the speech in the audio data obtained by the user device comprises synthetic speech.

Abstract: Methods, systems, and media for modifying the presentation of video content on a user device based on a consumption mode of the user device are provided. In some embodiments, a computer-implemented method comprises: receiving, from a user device, a request to present a video content item, wherein the video content item includes an audio portion; identifying a segment of the audio portion as including a particular type of audio content (e.g., silent content, music content, speech content, theatrical content, non-musical content, etc.); determining a consumption mode associated with the user device; generating a modified video content item based on the consumption mode associated with the user device by removing a portion of the video content item that corresponds to the segment of the audio portion identified as including the particular type of audio content; and causing the modified video content item to be presented on the user device.

Abstract: Methods, systems and apparatus for producing quantum circuits with low T gate counts. In one aspect, a method for performing a temporary logical AND operation on two control qubits includes the actions of obtaining an ancilla qubit in an A-state; computing a logical-AND of the two control qubits and storing the computed logical-AND in the state of the ancilla qubit, comprising replacing the A-state of the ancilla qubit with the logical-AND of the two control qubits; maintaining the ancilla qubit storing the logical-AND of the two controls until a first condition is satisfied; and erasing the ancilla qubit when the first condition is satisfied.

Abstract: A data parallel pipeline may specify multiple parallel data objects that contain multiple elements and multiple parallel operations that operate on the parallel data objects. Based on the data parallel pipeline, a dataflow graph of deferred parallel data objects and deferred parallel operations corresponding to the data parallel pipeline may be generated and one or more graph transformations may be applied to the dataflow graph to generate a revised dataflow graph that includes one or more of the deferred parallel data objects and deferred, combined parallel data operations. The deferred, combined parallel operations may be executed to produce materialized parallel data objects corresponding to the deferred parallel data objects.