Abstract: Various example embodiments described herein provide for systems, methods, devices, instructions, and the like for using AI model cascades to filter data on a data system, such as a database system, based on an artificial intelligence model prompt (e.g., user-provided prompt). In particular, various example embodiments enable a database system to use cascaded AI models and adaptive bounds to optimize data filtering operations based on an artificial intelligence model prompt (also referred to herein as just a prompt) while balancing computational cost and accuracy, which can be useful in processing large-scale data queries.

Abstract: The subject technology receives first party training data provided by an end-user of a baseline machine learning model. The subject technology determines a first set of common features based on the first party training data. The subject technology receives, from at least one data source. The subject technology determines a second set of common features based on the set of datasets. The subject technology trains, using the first set of common features and the second set of common features, a second machine learning model, the second machine learning model incorporating additional training data from the external data supplier during training compared to the baseline machine learning model. The subject technology generates a boosted machine learning model based at least in part on the training, the boosted machine learning model comprising the trained second machine learning model.

Abstract: Embodiments of the present disclosure provide techniques for classification with automated model selection, tuning, and training. A processing device receives, from a client, a data query referencing an input data set of a database associated with a virtual warehouse. The processing device allocates an amount of memory of the virtual warehouse to be used to train a machine learning (ML) model based on the input data set and a peak memory estimate, where the peak memory estimate is based on a heuristic. The processing device trains, based on the input data set and the data query, the ML model in the virtual warehouse using the amount of memory.

Abstract: Disclosed are techniques for anomaly detection in time series data using an ML model. An untrained time series forecasting machine learning (ML) model may be provided as part of a class that includes an anomaly detection function, a features module, and a target transform module. In response to the class being invoked, an instance of the time series forecasting ML model may be trained using training time series data specified in the invocation of the class. The trained instance of the forecasting ML model may be persisted in an anomaly detection object along with instances of the anomaly detection function, the features module, and the target transform module. In response to receiving a call to the anomaly detection object, performing anomaly detection on time series data specified in the call using at least the trained instance of the forecasting ML model and the instance of the anomaly detection function.

Abstract: Embodiments of the present disclosure describe systems, methods, and computer program products for redacting sensitive data within a database. An example method can include receiving a data query referencing unredacted data of a database, wherein the data query that is received comprises a value identifying a type of sensitive data to be redacted from the unredacted data, responsive to the data query, executing, by a processing device, a redaction operation to identify candidate sensitive data that matches the type of sensitive data to be redacted within the unredacted data of the database, and returning a redacted data set in which the candidate sensitive data that is provided is based on an authentication level utilized for execution of the redaction operation.

Abstract: Embodiments of the present disclosure describe systems, methods, and computer program products for redacting sensitive data within a database. An example method can include receiving a data query referencing unredacted data of a database, wherein the data query that is received comprises a value identifying a type of sensitive data to be redacted from the unredacted data, responsive to the data query, executing, by a processing device, a redaction operation to identify sensitive data that matches the type within the unredacted data of the database, and returning a redacted data set in which the sensitive data that matches the type is replaced or removed to the data query.

Abstract: Systems and methods for generating object references with selectable scopes are provided. The systems and methods perform operations including calling, by a first entity, a reference generator function using one or more arguments associated with a database object that the first entity is authorized to access according to a first set of access privileges, the one or more arguments comprising a scope definition that defines persistence of a reference. The operations include obtaining, from the reference generator function, a reference to the database object, the reference persisting according to the scope definition. The operations include passing the reference to a second entity to enable the second entity to perform one or more database operations on the database object according to a second set of access privileges derived from the first set of access privileges.

Abstract: Disclosed is a fast and accurate time series forecasting algorithm that eliminates the need for hyperparameter tuning. Time series data may be analyzed using a quadratic function to determine a quadratic trend prediction, which is removed from the time series data to generate first detrended time series data. A moving median of the time series data is determined and the moving median is removed from the time series data to generate second detrended time series data. An amplitude scaling factor is determined based on the second detrended time series data and the first detrended time series data is descaled using the amplitude scaling factor to generate descaled time series data. The descaled time series data is analyzed to determine a seasonal prediction and a time series forecast is generated based on the seasonal prediction, the quadratic trend prediction, and the amplitude scaling factor.

Abstract: Techniques for managing input and output error of a machine learning (ML) model in a database system are presented herein. Test data is generated from successive versions of a database system, the database system comprising a machine learning (ML) model to generate an output corresponding to a function of the database system The test data is used to train an error model to determine an error associated with the output of or an input to the ML model between the successive versions of the database system. In response to the ML model generating a first output based on a first input: the error model adjusts the first output when the error is associated with the output to the ML model and adjusts the first input when the error is associated with the input to the ML model.

Abstract: Using an attributes model of a time series forecasting model, determine a set of features based on time series data, the set of features including periodic components. The time series data may be divided into a set of segments. For each segment of the set of segments, a weight may be assigned using an age of the segment, resulting in a set of weighted segments of time series data. Using a trend detection model of the time series forecasting model, trend data from the set of weighted segments of time series data may be determined. A time series forecast may be generated by combining the set of features and the trend data.

Abstract: Systems and methods for managing input and output error of a machine learning (ML) model in a database system are presented herein. A set of test queries is executed on a first version of a database system to generate first test data, wherein the first version of the system comprises a ML model to generate an output corresponding to a function of the database system. An error model is trained based on the first test data and second test data generated based on a previous version of the system. The error model determines an error associated with the ML model between the first and previous versions of the system. The first version of the system is deployed with the error model, which corrects an output or an input of the ML model until sufficient data has been produced by the error model to retrain the ML model.

Abstract: An example method of tuning a machine learning operation can include receiving a data query comprising a reference to an input data set of a database, generating a plurality of hyperparameter sets based on the input data set, in response to receiving the data query, training a plurality of machine learning models using the plurality of hyperpararneter sets, selecting a first mathine learning model of the plurality of machine learning models based on an accuracy of an output of the first machine learning model, and in response to receiving the data query, returning the output of the first machine learning model.

Abstract: The subject technology receives first party training data provided by an end-user of a baseline machine learning model. The subject technology determines a first set of common features based on the first party training data. The subject technology receives, from at least one data source. The subject technology determines a second set of common features based on the set of datasets. The subject technology trains, using the first set of common features and the second set of common features, a second machine learning model, the second machine learning model incorporating additional training data from the external data supplier during training compared to the baseline machine learning model. The subject technology generates a boosted machine learning model based at least in part on the training, the boosted machine learning model comprising the trained second machine learning model.

Abstract: An example method of tuning a machine learning operation can include receiving a data query comprising a reference to an input data set of a database, generating a plurality of unique sets of hyperparameters by varying a hyperparameter value of each set of hyperparameters of the plurality of unique sets of hyperparameters based on the input data set, in response to receiving the data query, training a plurality of machine learning models using the input data set of the data query, each of the plurality of machine learning models configured according to a respective one of a plurality of unique sets of hyperparameters, selecting a first machine learning model of the plurality of machine learning models based on an accuracy of an output of the first machine learning model, and returning the output of the first machine learning model in response to the data query.

Abstract: Using an attributes model of a time series forecasting model, determine a set of features based on time series data, the set of features including periodic components. The time series data may be divided into a set of segments. For each segment of the set of segments, a weight may be assigned using an age of the segment, resulting in a set of weighted segments of time series data. Using a trend detection model of the time series forecasting model, trend data from the set of weighted segments of time series data may be determined. A time series forecast may be generated by combining the set of features and the trend data.

Abstract: Embodiments of the present disclosure describe systems, methods, and computer program products for redacting sensitive data within a database. An example method can include receiving a data query referencing unredacted data of a database, wherein the data query that is received comprises a value identifying a type of sensitive data to be redacted from the unredacted data, responsive to the data query, executing, by a processing device, a redaction operation to identify sensitive data that matches the type within the unredacted data of the database, and returning a redacted data set in which the sensitive data that matches the type is replaced or removed to the data query.

Abstract: An example method of tuning a machine learning operation can include receiving a data query comprising a reference to an input data set of a database, generating a plurality of unique sets of hyperparameters by varying a hyperparameter value of each set of hyperparameters of the plurality of unique sets of hyperparameters based on the input data set, in response to receiving the data query, training a plurality of machine learning models using the input data set of the data query, each of the plurality of machine learning models configured according to a respective one of a plurality of unique sets of hyperparameters, selecting a first machine learning model of the plurality of machine learning models based on an accuracy of an output of the first machine learning model, and returning the output of the first machine learning model in response to the data query.

Abstract: Systems and methods for managing input and output error of a machine learning (ML) model in a database system are presented herein. A set of test queries is executed on a first version of a database system to generate first test data, wherein the first version of the system comprises a ML model to generate an output corresponding to a function of the database system. An error model is trained based on the first test data and second test data generated based on a previous version of the system. The error model determines an error associated with the ML model between the first and previous versions of the system. The first version of the system is deployed with the error model, which corrects an output or an input of the ML model until sufficient data has been produced by the error model to retrain the ML model.

Abstract: A processing device may analyze a set of time series data using a time series forecasting model comprising an attributes model and a trend detection model. The attributes model may comprise a modified gradient boosting decision tree (GBDT) based algorithm. Analyzing the set of time series data comprises determining a set of features of the set of time series data, the set of features including periodic components as well as arbitrary components. A trend of the set of time series data may be determined using the trend detection model and the set of features and the trend may be combined to generate a time series forecast.

Abstract: Embodiments of the present disclosure describe systems, methods, and computer program products for redacting sensitive data within a database. An example method can include receiving a data query referencing unredacted data of a database, responsive to the data query, executing, by a processing device, a redaction operation to identify sensitive data within the unredacted data of the database, and returning a redacted data set in which the sensitive data is replaced or removed to the data query.