Search for: All records

Successful supervised learning models rely on predictive features, which rarely come from a single dataset. As a result, relevant datasets need to be integrated before training the actual model. This raises one natural question: \textit{``how can one efficiently search for predictive features from relevant datasets for integration with responsible AI guarantees?"}. This paper formalizes the question as the \textit{data augmentation search problem} with an objective of minimizing the search latency. We propose \sys, an interactive system that intakes a supervised learning task and searches for a set of join-compatible datasets that optimally improve the performance of the task. Specifically, \sys manages a corpus of relational datasets, uses linear regression as a \textit{proxy model} to evaluate augmentation candidates, and applies \textit{factorized machine learning} to accelerate model training and evaluation algorithmically. Furthermore, \sys leverages system and hardware optimizations to maximize parallelism across augmentation searches. These allow \sys to search for a good augmentation plan over 1 million datasets with a latency of $1.4$ seconds. 

Recent platforms utilize ML task performance metrics, not metadata keywords, to search large data corpus. Requesters provide an initial dataset, and the platform searches for additional datasets that augment---join or union---requester's dataset to most improve the model (e.g., linear regression) performance. Although effective, current task-based data searches are stymied by (1) high latency which deters users, (2) privacy concerns for regulatory standards, and (3) low data quality which provides low utility. We introduce Mileena, a fast, private, and high-quality task-based dataset search platform. At its heart, Mileena is built on pre-computed semi-ring sketches for efficient ML training and evaluation. Based on semi-ring, we develop a novel Factorized Privacy Mechanism that makes the search differentially private and scales to arbitrary corpus sizes and numbers of requests without major quality degradation. We also demonstrate the early promise in using LLM-based agents for automatic data transformation and applying semi-rings to support causal discovery and treatment effect estimation. 

Although dominant for tabular data, ML libraries that train tree models over normalized databases (e.g., LightGBM, XGBoost) require the data to be denormalized as a single table, materialized, and exported. This process is not scalable, slow, and poses security risks. In-DB ML aims to train models within DBMSes to avoid data movement and provide data governance. Rather than modify a DBMS to support In-DB ML, is it possible to offer competitive tree training performance to specialized ML libraries...with only SQL? We present JoinBoost, a Python library that rewrites tree training algorithms over normalized databases into pure SQL. It is portable to any DBMS, offers performance competitive with specialized ML libraries, and scales with the underlying DBMS capabilities. JoinBoost extends prior work from both algorithmic and systems perspectives. Algorithmically, we support factorized gradient boosting, by updating theYvariable to the residual in thenon-materialized join result.Although this view update problem is generally ambiguous, we identifyaddition-to-multiplication preserving, the key property of variance semi-ring to supportrmsethe most widely used criterion. System-wise, we identify residual updates as a performance bottleneck. Such overhead can be natively minimized on columnar DBMSes by creating a new column of residual values and adding it as a projection. We validate this with two implementations on DuckDB, with no or minimal modifications to its internals for portability. Our experiment shows that JoinBoost is 3× (1.1×) faster for random forests (gradient boosting) compared to LightGBM, and over an order of magnitude faster than state-of-the-art In-DB ML systems. Further, JoinBoost scales well beyond LightGBM in terms of the # features, DB size (TPC-DS SF=1000), and join graph complexity (galaxy schemas). 

Modern database management systems employ sophisticated query optimization techniques that enable the generation of efficient plans for queries over very large data sets. A variety of other applications also process large data sets, but cannot leverage database-style query optimization for their code. We therefore identify an opportunity to enhance an open-source programming language compiler with database-style query optimization. Our system dynamically generates execution plans at query time, and runs those plans on chunks of data at a time. Based on feedback from earlier chunks, alternative plans might be used for later chunks. The compiler extension could be used for a variety of data-intensive applications, allowing all of them to benefit from this class of performance optimizations.