Abstract: Methods and systems described herein for validating machine learning models in federated machine learning model environments. More specifically, the methods and systems relate to unloading training and validation techniques to client devices using newly collected data to improve accuracy of federated machine learning models.

Abstract: Systems and methods for an artificial intelligence model architecture that involves a first artificial intelligence model trained to map a plurality of entities to ranked documentation from a documentation source, and a second artificial intelligence model that comprises a language model trained to generate an additional query to run on the plurality of documents from the documentation source. By training the second model to generate additional queries as entities and/or links are discovered, the system may quickly and efficiently determine links and/or potential resolutions as well as received feedback thereon.

Abstract: Methods and systems are described herein for protecting client data while training machine learning models. The system may transmit, to client devices, simple models to be trained on a respective client device to generate predictions based on a respective subset of respective client data of the respective client device. The system may receive the trained simple models from the client devices. The system may input, into an ensemble model including the simple models, an unlabeled synthetic dataset. This may cause the ensemble model to aggregate a set of predictions generated by each simple model to generate labels for the unlabeled synthetic dataset. The system may then input, into a new model, the unlabeled synthetic dataset and the labels to train the new model to predict the labels for the unlabeled synthetic dataset.

Abstract: A method includes sending a data selection parameter to a client computing device, wherein the client computing device stores a distributed instance of a machine learning model, and the machine learning model includes at least a first subset of parameters and a second subset of parameters. In response to a determination that a transmission criterion is satisfied, the client computing device is caused to report the first subset of parameters based on the data selection parameter. The method further comprises obtaining the parameters of the first subset of parameters of the distributed instance from the client computing device and updating the federated learning model based on the first subset of parameters of the distributed instance from the client computing device.

Abstract: Methods and systems for generating federated learning models. In some aspects, the system receives, from each client device, user data profiles that are anonymized with respect to users associated with user data stored locally at a client device. The system processes the user data profiles to generate a plurality of clusters. For each cluster, the system transmits, to one or more client devices corresponding to a cluster, a first instruction to train a machine learning model on user data corresponding to user data profiles included in the cluster and a second instruction to validate the machine learning model with respect to user data corresponding to one or more clusters of the plurality of clusters other than the cluster to generate a prediction accuracy metric. The system determines, from the plurality of clusters, a first cluster based on associated prediction accuracy metrics.

Abstract: Methods and systems are described herein for novel uses and/or improvements to artificial intelligence applications for data-sparse environments. As one example, methods and systems are described herein for overcoming the problem of determining an appropriate digital asset to recommend to a user in real-time based on training data that may be shared between multiple intents (e.g., as would be found in responses to textual, verbal, and/or other real-time communications). In particular, the methods and systems overcome this technical problem by using an artificial intelligence trained to identify specific verbiage of a user to display digital assets corresponding to a user's intent thereby increasing productivity, organization, and collaboration, and reducing frustration.

Abstract: Methods and systems are described for novel uses and/or improvements to federated learning. As one example, methods and systems are described for improving the applicability of federated learning across various applications and increasing the efficiency of training a global model through federated learning. As another example, methods and systems are described for ensuring comprehensive training data is available to models assigned by the federated learning server. Additionally, methods and systems are described for improving the rate of training a global model through federated learning.

Abstract: The systems and methods use a divergent artificial intelligence architecture. To summarize contextually relevant information on user-selected concepts despite the lack of adequate training data on each user, the system may use an architecture in which a first model is trained to map a plurality of user interactions to inputted concept identifiers to determine user relationship matrices whereas a second model is trained to generate contextually relevant summaries on inputted concepts and inputted user relationship matrices. For example, by using this divergent structure, the system does not need to be specifically trained to generate contextually relevant summaries based on each individual user's relationship to a given concept.

Abstract: Methods and systems are described herein for protecting client data while training machine learning models. The system may transmit, to client devices, simple models to be trained on a respective client device to generate predictions based on a respective subset of respective client data of the respective client device. The system may receive the trained simple models from the client devices. The system may input, into an ensemble model including the simple models, an unlabeled synthetic dataset. This may cause the ensemble model to aggregate a set of predictions generated by each simple model to generate labels for the unlabeled synthetic dataset. The system may then input, into a new model, the unlabeled synthetic dataset and the labels to train the new model to predict the labels for the unlabeled synthetic dataset.

Abstract: Systems and methods for an artificial intelligence model architecture that involves a first artificial intelligence model trained to map a plurality of entities to ranked documentation from a documentation source, and a second artificial intelligence model that comprises a language model trained to generate an additional query to run on the plurality of documents from the documentation source. By training the second model to generate additional queries as entities and/or links are discovered, the system may quickly and efficiently determine links and/or potential resolutions as well as received feedback thereon.

Abstract: Methods and systems are described herein for protecting client data while training machine learning models. The system may transmit, to client devices, simple models to be trained on a respective client device to generate predictions based on a respective subset of respective client data of the respective client device. The system may receive the trained simple models from the client devices. The system may input, into an ensemble model including the simple models, an unlabeled synthetic dataset. This may cause the ensemble model to aggregate a set of predictions generated by each simple model to generate labels for the unlabeled synthetic dataset. The system may then input, into a new model, the unlabeled synthetic dataset and the labels to train the new model to predict the labels for the unlabeled synthetic dataset.

Abstract: Methods and systems for scalable dataset content embedding for improved searchability. For example, the system may retrieve a first dataset from a first data source. The system may generate a first data profile of the first dataset. The system may generate a latent index of the first data profile based on processing the first data profile using a first embedding algorithm. The system may receive, via a user interface, a first request for a first text string. The system may generate an embedded request corresponding to the first request based on processing the first text string using the first embedding algorithm. The system may process the embedded request using the latent index. The system may generate for display, in the user interface, a result based on processing the embedded request using the latent index.

Abstract: Methods and systems described herein relate to the creation of a digital repository of artificial intelligence models that allows users to determine their individual fairness metric. More specifically, the methods and systems provide this digital repository by storing it on a blockchain network and tracking any changes made to the model and/or its fairness metric.

Abstract: Systems and methods for generating synthetic data samples. In some aspects, a system accesses first and second user data samples corresponding to a first and second period of time, respectively, and generates a first and second profile based on the user data samples, wherein the profiles comprise parameters representing metadata associated with corresponding user data samples. The system generates a new profile corresponding to an intermediary period of time between a first and second period of time, wherein the new profile comprises intra-profile and inter-profile parameters. The system determines (1) a value for each intra-profile parameter based on values of intra-profile parameters of the first and second profile and (2) the value for each inter-profile parameter based on a predicted value output of a model trained on a plurality of data profiles over time and generates synthetic data samples based on values of the inter-profile and intra-profile parameters.

Abstract: Systems and methods for novel uses and/or improvements to data labeling applications, particularly data labeling applications involving sensitive data. As one example, systems and methods are described herein for preventing sensitive data leakage, using weak learner libraries, during label propagation.

Abstract: In some implementations, a device may obtain data indicating reparations issued by an entity that uses artificial intelligence to provide artificial intelligence outputs in connection with users, the reparations being issued for one or more of the artificial intelligence outputs being erroneous. The device may determine, using a machine learning model, an artificial intelligence reparation characterization for the entity. The artificial intelligence reparation characterization determined using the machine learning model may be indicative of an amount of reparations predicted for the entity in connection with uses of artificial intelligence by the entity. The machine learning model may be trained to determine the artificial intelligence reparation characterization based on the data. The device may transmit information indicating the artificial intelligence reparation characterization.

Abstract: Methods and systems for scalable dataset content embedding for improved searchability. For example, the system may retrieve a first dataset from a first data source. The system may generate a first data profile of the first dataset. The system may generate a latent index of the first data profile based on processing the first data profile using a first embedding algorithm. The system may receive, via a user interface, a first request for a first text string. The system may generate an embedded request corresponding to the first request based on processing the first text string using the first embedding algorithm. The system may process the embedded request using the latent index. The system may generate for display, in the user interface, a result based on processing the embedded request using the latent index.

Abstract: Systems and methods for novel uses and/or improvements to data labeling applications, particularly data labeling applications involving sensitive data. As one example, systems and methods are described herein for preventing sensitive data leakage, using weak learner libraries, during label propagation.

Abstract: A method and related system for storing data based on predictions includes obtaining, from a client device, a first set of update data and a time-related prediction generated by a client-side version of a machine learning model. The method further includes determining that the time-related prediction is associated with a first data store of a plurality of data stores comprising the first data store and a second data store. The method further includes updating a record in the first data store based on the first set of update data in response to a determination that the time-related prediction is associated with the first data store. The method further includes updating the record in the first data store based on a second set of update data obtained after obtaining the first set of update data.

Abstract: A method and related system for visualizing search result information based on sentiment includes operations to retrieve sentiment scores associated with retrieved records associated with display scores and comprising values in a latent space, determining a centering sentiment based on the sentiment scores, and determining sentiment ranges and subranges based on the centering sentiment. The method further includes selecting a first record based on display scores of records within the first subrange, sending data comprising the sentiment range to a client device that causes the client device to present a visualization of a region bounded by the sentiment range. The visualization includes a first shape representing the first record positioned in a first subregion associated with a positive sentiment and a second shape representing a record within the second subrange positioned in a second subregion associated with a negative sentiment.