Abstract: System, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, generate a prediction based on a query. A query is provided to a deep machine learning (ML) model. The deep ML model generates a plurality of query projection embeddings by projecting the query into each of a plurality of different query embedding spaces and generates the prediction based at least on the plurality of query projection embeddings. Each of a plurality of query projection embedding layers of the deep ML may generate a corresponding one of the query projection embeddings by applying a hash function associated with the query projection layer to the query to generate a vector representation of the query, and applying a set of weights associated with the query projection layer to the vector representation to generate a query projection embedding in the plurality of query projection embeddings.

Abstract: Embodiments present techniques for determining a list of recommended items in response to a user query. An embodiment can determine a first ordered list of items including a plurality of items stored by a content platform. Based on a reward discount parameter, a first total discounted future reward for the first ordered list of items can be determined. Based on a risk discount parameter, a first risk estimate for the first ordered list of items can be determined. Similarly, a second ordered list of items can have a second total discounted future reward and a second risk estimate. The second ordered list of items can be the list of recommended items when the second total discounted future reward is larger than or equal to the first total discounted future reward, and the second risk estimate is less than or equal to the first risk estimate.

Abstract: Aspects of the disclosure are directed to performing hierarchical queries based on summary reporting from an application programming interface (API) for digital content estimation. Performing the hierarchical queries can include denoising API outputs while ensuring consistency across different levels of a hierarchy. Performing the hierarchical queries can further include optimizing a privacy budget across different levels of the hierarchy.

Abstract: Disclosed are system, method and/or computer program product embodiments that retrieve items for a user based on a query using a two-tower deep machine learning model. An example embodiment provides input to a context tower, wherein the input includes the query and one or more of a query embedding corresponding to the query or a graph user embedding corresponding to the user. The context tower generates a context embedding in a vector space based on the input. The model determines a measure of similarity between the context embedding and each of a plurality of item embeddings in the vector space that are generated by an item tower and represent a plurality of candidate items. A relevancy score is calculated for each candidate item based on the measure of similarity between the context embedding and the corresponding item embedding. The relevancy scores are used for item retrieval and/or ranking.

Abstract: Embodiments present techniques for determining a list of recommended items in response to a user query. An embodiment can determine a first ordered list of items including a plurality of items stored by a content platform. Based on a reward discount parameter, a first total discounted future reward for the first ordered list of items can be determined. Based on a risk discount parameter, a first risk estimate for the first ordered list of items can be determined. Similarly, a second ordered list of items can have a second total discounted future reward and a second risk estimate. The second ordered list of items can be the list of recommended items when the second total discounted future reward is larger than or equal to the first total discounted future reward, and the second risk estimate is less than or equal to the first risk estimate.

Abstract: The disclosure generally describes methods, software, and systems for integration of attribution reports and auxiliary data sources. Data including aggregated summary reports and event level reports is received from a first system. Additional raw affirmative action data related to the aggregated summary reports and event level reports is received from a second system. A matrix is created with rows for interaction events and columns for raw affirmative action data. Denoised aggregated counts and denoised event counts are determined from the received reports. The matrix fields are optimized by resolving conflicts between raw counts, denoised aggregated counts, and denoised event counts.

Abstract: Disclosed herein are system, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for ranking a plurality of content items for presentation to a user. An embodiment generates a ranking score for each content item by: providing input to a deep machine learning (ML) model, the input including at least one or more query features and one or more content item features, determining, by the deep ML model and based at least on the input, a first probability of a first type of interaction between the user and the content item and a second probability of a second type of interaction between the user and the content item, and calculating the ranking score for the content item based at least on the first and second probabilities. An embodiment ranks the content items for presentation based on the ranking score associated with each content item.

Abstract: A method is described and includes receiving by a query planner agent of a query response system a user query comprising a request for a response; determining a type of the received user query; identifying one of a plurality of response modules comprising the query response system based on the determined type of the received user query; and forwarding the received user query to the identified one of the plurality of response modules. In example embodiments, the determined type of the received user query comprises one of a lexical query, a categorical query, an explanatory query, and a multistep query, each of which is routed to a different response module.

Abstract: A dual-screen computing device includes two separate displays that are coupled to an interconnecting hinge. A hinge detector detects movement or position of the hinge, and the positions of the displays may be determined based on the hinge movement or position. The positions of the displays relative to each other may then be used to determine which mode of operation the dual-screen computing device is operating (e.g., tent mode, open, closed, etc.). Additionally, the dual-screen computing device may include various sensors that detect different environmental, orientation, location, and device-specific information. Applications are configured to operate differently based on the mode of operation and, optionally, the sensor data detected by the sensors.

Abstract: Crest factor reduction circuitry includes: a peak neighborhood analyzer; peak detection circuitry and a controller. The peak neighborhood analyzer is configured to: receive an input signal; analyze the input signal to determine whether a peak larger than a target threshold is expected within an interval; and provide a first control signal responsive to determining that a peak larger than the target threshold is expected within the interval. The controller is configured to: receive the first control signal; and gate a clock or data to the peak detection circuitry responsive to the first control signal.

Abstract: Disclosed are system, method and/or computer program product embodiments that retrieve items for a user based on a query using a two-tower deep machine learning model. An example embodiment provides input to a context tower, wherein the input includes the query and one or more of a query embedding corresponding to the query or a graph user embedding corresponding to the user. The context tower generates a context embedding in a vector space based on the input. The model determines a measure of similarity between the context embedding and each of a plurality of item embeddings in the vector space that are generated by an item tower and represent a plurality of candidate items. A relevancy score is calculated for each candidate item based on the measure of similarity between the context embedding and the corresponding item embedding. The relevancy scores are used for item retrieval and/or ranking.

Abstract: Aspects of the disclosed technology provide solutions for extracting subgraph patterns in graph-structured data and encoding them as embeddings using a graph neural network (GNN). In some aspects, a process of the disclosed technology can include steps for receiving an input graph comprising a plurality of nodes and edges, the input graph representing relationships among a plurality of entities, parameterizing a graph neural network model based on a set of pattern graphs, and identifying, for at least a portion of the nodes in the input graph, rooted homomorphisms between the pattern graphs and local subgraphs rooted at the respective nodes. Systems and machine-readable media are also provided.

Abstract: Disclosed herein are system, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for ranking a plurality of content items for presentation to a user. An embodiment generates a ranking score for each content item by: providing input to a deep machine learning (ML) model, the input including at least one or more query features and one or more content item features, determining, by the deep ML model and based at least on the input, a first probability of a first type of interaction between the user and the content item and a second probability of a second type of interaction between the user and the content item, and calculating the ranking score for the content item based at least on the first and second probabilities. An embodiment ranks the content items for presentation based on the ranking score associated with each content item.

Abstract: Large language models can receive a prompt and generate responses having natural language and/or data structures as sequence of tokens. Some responses may have a few dozen tokens. The speed of response of large language models can be directly proportional to how many tokens are being generated. Rather than producing many tokens, it is possible to fine-tune a large language model to generate responses in an encoded output format. A response can have one or more encoded values that can indicate the same information as a natural language and/or structured data response. The response may include only a single or few tokens. The speed of response of a large language model operating as an encoder would be faster.

Abstract: Crest factor reduction circuitry includes: a peak neighborhood analyzer; peak detection circuitry and a controller. The peak neighborhood analyzer is configured to: receive an input signal; analyze the input signal to determine whether a peak larger than a target threshold is expected within an interval; and provide a first control signal responsive to determining that a peak larger than the target threshold is expected within the interval. The controller is configured to: receive the first control signal; and gate a clock or data to the peak detection circuitry responsive to the first control signal.

Abstract: A method is described and includes obtaining a list of utterances comprising captions from an item of content; computing sentence transformer embeddings for each of the utterances; dividing the utterances into sentences and extracting a sentence embedding for each sentence; computing a semantic similarity between adjacent sentences; and merging the adjacent sentences into a block comprising a segment if the semantic similarity between the adjacent sentences is greater than a predetermined threshold.

Abstract: Embodiments present techniques for determining a list of recommended items in response to a user query. An embodiment can determine a first ordered list of items including a plurality of items stored by a content platform. Based on a reward discount parameter, a first total discounted future reward for the first ordered list of items can be determined. Based on a risk discount parameter, a first risk estimate for the first ordered list of items can be determined. Similarly, a second ordered list of items can have a second total discounted future reward and a second risk estimate. The second ordered list of items can be the list of recommended items when the second total discounted future reward is larger than or equal to the first total discounted future reward, and the second risk estimate is less than or equal to the first risk estimate.

Abstract: Retrieving content items in response to a query in a way that increases user satisfaction and increases chances of users consuming a retrieved content item is not trivial. One retrieval strategy may include dividing the content items into buckets according to a dimension about the content items and retrieving a top K number of items from different buckets to balance semantic affinity and the dimension. Choosing an optimal K for different buckets for a given query can be a challenge. Reinforcement learning can be used to train and implement an agent model that can choose the optimal K for different buckets.

Abstract: Pre-trained large language models may be trained on a large data set which may not necessarily align with specific tasks, business goals, and requirements. Pre-trained large language models can solve generic semantic relationship or question-answering type problems but may not be suited for content item retrieval or recommendation of content items that are semantically relevant to a query. It is possible to build a machine learning model while using transfer learning to learn from pre-trained large language models. Training data can significantly impact the performance of machine learning models, especially machine learning models developed using transfer learning. The training data can impact a model's performance, generalization, fairness, and adaptation to specific domains. To address some of these concerns, a popularity bucketing strategy can be implemented to debias training data. Optionally, an ensemble of models can be used to generate diverse training data.

Abstract: An example apparatus described herein to implement cancellation pulse generation includes a first memory storing first subsets of data samples of a single pulse cancellation waveform. The example apparatus includes a second memory storing second subsets of data samples of the single pulse cancellation waveform, the second subsets including different data samples of the single pulse cancellation waveform than the first subsets. The example apparatus includes first circuitry coupled to the first memory and to the second memory in parallel. The example apparatus includes a plurality of buffers. The example apparatus includes second circuitry coupled to the plurality of buffers.