Abstract: In some embodiments, a system-on-chip, includes a central processing unit (CPU); an accelerator coupled to the CPU via a first die-to-die interconnect; and uniform memory coupled to the CPU via a second die-to-die interconnect. In some embodiments, in order to prevent use of accelerator memory for processing operations by the accelerator, the accelerator utilizes a uniform memory access tunneling system located in the accelerator to tunnel a high-level interconnect protocol associated with the second die-to-die interconnect to a die-to-die interconnect protocol associated with the first die-to-die interconnect, the uniform memory access tunneling system being configured to allow access to the uniform memory using a shared address space.

Abstract: In one embodiment, a first device may receive, from a second device, a reference landmark map identifying locations of facial features of a user of the second device depicted in a reference image and a feature map, generated based on the reference image, representing an identity of the user. The first device may receive, from the second device, a current compressed landmark map based on a current image of the user and decompress the current compressed landmark map to generate a current landmark map. The first device may update the feature map based on a motion field generated using the reference landmark map and the current landmark map. The first device may generate scaling factors based on a normalization facial mask of pre-determined facial features of the user. The first device may generate an output image of the user by decoding the updated feature map using the scaling factors.

Abstract: An audio processing system includes: a receiver configured to receive the original audio data; a processor configured to execute the instructions stored in the memory to cause the audio processing system to: separate a background noise audio data, a first speaker audio data, and a second speaker audio data; recognize first speaker speech, convert the first speaker speech to first speaker text, translate the first speaker text to a second language text, and convert the second language text to a second speech; recognize second speaker speech, convert the second speaker speech to second speaker text, translate the second speaker text to the second language text, and convert the second language text of the second speaker to a second speech for the second speaker; and generate encoded audio data; and a transmitter configured to transmit the encoded audio data to a content user device.

Abstract: In one embodiment, a method includes, by one or more computing devices, receiving, from a client system of a first user of an online social network, a text post inputted by the first user, parsing the text post to identify a query associated with the text post, sending, to the client system, instructions for presenting a confirmation prompt requesting confirmation of the identified query from the first user, receiving, from the client system, a confirmation of the identified query from the first user, generating, in response to receiving the confirmation, a recommendation list responsive to the query, wherein the recommendation list comprises references to one or more objects referenced in one or more prior comments associated with one or more prior posts of the online social network associated with the query, and sending, to the client system, instructions for presenting the recommendation list to the first user.

Abstract: Systems, apparatuses and methods provide technology that determines that a primary slice of a plurality of slices of a communication network is overloaded. The plurality of slices includes a backup slice. The technology assigns user equipment to operate with the backup slice in response to the primary slice being overloaded. The technology identifies that a trigger has occurred, wherein the trigger is associated with one or more of the user equipment, the backup slice or the primary slice, and re-assigns the user equipment to the primary slice from the backup slice so that the user equipment operates with the primary slice.

Abstract: Systems, apparatuses and methods provide technology that identifies a first post that is submitted to a first group of a social network. The technology identifies that the first post is a cross-pollination candidate, identifies a second group of the social network, generates a first vector that is to represent one of the first post or the first group, generates a second vector that is to represent the second group, determines whether the second group matches a cross-pollination criteria based on a comparison of the first vector to the second vector, and determines whether to automatically generate a second post based on the first post, and submit the second post to the second group based on whether the second group matches the cross-pollination criteria.

Abstract: Exemplary embodiments relate to the automatic generation of captions for visual media, including photos, photo albums, non-live video, and live video. The visual media may be analyzed to determine contextual information (such as location information, people and objects in the video, time, etc.). A system may integrate this information with information from the user's social network and a personalized language model built using public-facing language from the user. The personalized language model captures the user's way of speaking to make the generated captions more detailed and personalized. The language model may account for the context in which the video was generated. The captions maybe used to simplify and encourage content generation, and may also be used to index visual media, rank the media, and recommend the media to users likely to engage with the media.

Abstract: In one embodiment, a method includes receiving, from a client system of a user, a user input comprising a plurality of n-grams, parsing the user input to identify one or more overall intents, hidden intents, and slots associated with the one or more n-grams, wherein at least one of the hidden intents is non-resolvable for being associated with partial slot information corresponding to an n-gram that has not been resolved to a particular entity identifier, wherein the partial slot information is associated with two more entity identifiers of two or more entities, respectively, sending, to the client system, instructions for prompting the user to select one of the entities to be associated with the non-resolvable hidden intent, resolving the non-resolvable hidden intent based on the entity identifier of the entity selected by the user, and generating a response to the user input based on the resolved hidden intent.

Abstract: In one embodiment, a method includes receiving a user input including an ambiguous mention by a user from the user's client system, analyzing the user input to identify intents and slots based on a natural-language understanding module, where at least one slot is an ambiguous slot corresponding to the ambiguous mention, identifying candidate entities corresponding to the ambiguous slot, ranking the candidate entities based on the user's user profile data, sending to the user's client system instructions for prompting the user to select a candidate entity from a set of candidate entities having a rank greater than a threshold rank, where the selected candidate entity is resolved to the ambiguous slot, and storing the selected candidate entity together with the ambiguous mention as a mention-entity pair as part of the user profile data responsive to receiving the selection from the user.

Abstract: A system including a machine-learning accelerator (MLA) hardware comprising computation-control units that each have a programmable dependency matrix; and a compiler computing module configured to generate, based on a machine-learning model, dependency instructions indicating dependencies between the computation-control units; wherein the computation-control units include at least: a first computation-control unit configured to generate, after completion of a first operation, a synchronization token representing the completion of the first operation, the synchronization token specifying a recipient identifier for an intended recipient computation-control unit of the synchronization token; a second computation-control unit configured to: configure the programmable dependency matrix of the second computation-control unit according to the dependency instructions to include dependency conditions for performing operations; receive the synchronization token based on the recipient identifier; update a dependency sta

Abstract: An image captured by an imaging device in low light conditions may cause the captured image to have poor contrast, blurring, and otherwise not display one or more objects in the image clearly. According to various aspects, systems and methods are provided for enhancing images that are captured in low light conditions.

Abstract: Systems, apparatuses and methods provide technology that identifies historical data for a first user, where the historical data identifies previous posts by the first user. The technology identifying an event. The technology determines an affinity score for the first user based on the historical data and the event, determines whether the affinity score meets a threshold, and if the affinity score meets the threshold, determines one or more of a phrase or an effect to present to the first user.

Abstract: Systems, apparatuses and methods provide technology that identifies first characteristics of a first risk associated with a system and applies a locality sensitive hashing process to the first characteristics to map the first characteristics to first buckets of a plurality of buckets. The technology further generates a first vector based on the first characteristics being mapped to the first buckets, and identifies a mitigation plan to at least partially mitigate the first risk based on the first vector.

Abstract: Systems, apparatuses and methods provide technology that generates first data, stores the first data in a data storage of a first computing device to generate first stored data, transmits the first data as first test data along a first integrity path comprising at least one first hardware device. The technology further receives, with the first computing device, the first test data from the first integrity path and compares the first stored data to the received first test data to determine if a first data path error exists in the first integrity path.

Abstract: A system and method for Crest Factor Reduction (CFR) pulse cancellation (PC) in a single carrier environment or multicarrier telecommunication environment may enable obtaining better signal quality while maintaining Peak to Average Power Ratio (PAPR) and/or power efficiency. A PC signal is generated by multiplying a truncated sinc signal with another window signal. The bandwidth of the PC signal may be greater than the bandwidth of the corresponding carrier signal center. The center frequency of the PC signal may be offset with respect to the center frequency for each given carrier in the multi-carrier scenario to fix the edge effect signal quality/interference problems.

Abstract: An online system customizes video conversations between users of the online system. During a video conversation, the online system presents a composite view to the participating users. The composite view may include visual representations of the users, a background graphic, or other types of graphics such as masks and props that the users can wear or interact with in the environment of the video conversation. The visual representations may be generated based on a live video feed of the users or include avatars of the users. The online system can determine the graphics based on information about the users. For instance, the online system determines a background graphic showing a location that the users have each visited. Upon viewing the background graphic, the users may be encouraged to interact with the background graphic or other graphics included in the composite view, which can promote an engaging video conversation experience.

Abstract: In one embodiment, a method includes presenting a suggestion to a user of a head-mounted device by the head-mounted device via an assistant xbot during a dialog session between the user and the assistant xbot, wherein the suggestion is associated with a plurality of actions to be performed by an assistant system associated with the assistant xbot, accessing signals from inertial measurement unit (IMU) sensors of the head-mounted device by the head-mounted device during the dialog session, determining a head gesture performed by the user during the dialog session by an on-device head-gesture detection model and based only on the signals from the IMU sensors, and executing a first action from multiple actions by the assistant system executing on the head-mounted device, wherein the first action is selected based on the determined head gesture during the dialog session.

Abstract: A computer-implemented method for protecting identity metrics may include (i) receiving, by a computing device and from an originating device, an identity metric encrypted by an identity metric protection module on the originating device, where the identity metric includes biometric data about an operator of the originating device, (ii) requesting, by an identity metric protection module on the computing device, a decryption key from an identity metric protection server, (iii) receiving, by the identity metric protection module, the decryption key from the identity metric protection server, (iv) decrypting, by the identity metric protection module, the identity metric with the decryption key received from the identity metric protection server, and (v) providing the decrypted identity metric to an application on the computing device that uses the identity metric to convey a communication from the operator of the originating device. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: Technology for identifying an object of interest includes obtaining object embeddings for a plurality of objects in an image, obtaining text embeddings for text associated with the image, determining, for each of the plurality of objects, a similarity score via a similarity model based on the text embeddings and the object embeddings, while bypassing use of bounding box coordinates, and selecting the object having the highest similarity score as the object of interest. In another example, technology for identifying an object of interest includes obtaining object embeddings for a plurality of objects in an image, obtaining text embeddings and text identifiers for text associated with the image, generating, via a single transformer encoder, a set of CLS embeddings based on the text embeddings and the object embeddings, and determining, via a neural network, the object of interest based on the CLS embeddings.

Abstract: Systems, methods, and non-transitory computer-readable media can be configured to determine a life story associated with a first user of a content sharing system. The life story can comprise a plurality of content items associated with the first user. The plurality of content items can be organized in chronological order. A determination is made that a second user has accessed the life story associated with the first user. The plurality of content items included in the life story are provided for presentation in a graphical user interface, wherein each content item is presented individually based on the chronological order.