The integration of hardware accelerators has significantly advanced the capabilities of modern recommendation systems, enabling the exploration of complex ranking paradigms previously deemed impractical. However, the GPU-based computational costs present substantial challenges. In this paper, we demonstrate our development of an efficiency-driven approach to explore these paradigms, moving beyond traditional reliance on native PyTorch modules. We address the specific challenges posed by ranking models’ dependence on categorical features, which vary in length and complicate GPU utilization. We introduce Jagged Feature Interaction Kernels, a novel method designed to extract fine-grained insights from long categorical features through efficient handling of dynamically sized tensors. We further enhance the performance of attention mechanisms by integrating Jagged tensors with Flash Attention. Our novel Jagged Flash Attention achieves up to 9 × speedup and 22 × memory reduction compared to dense attention. Notably, it also outperforms dense flash attention, with up to 3 × speedup and 53% more memory efficiency. In production models, we observe 10% QPS improvement and 18% memory savings, enabling us to scale our recommendation systems with longer features and more complex architectures.

Heterogeneous information networks (HINs), which are typed graphs with labeled nodes and edges, have attracted tremendous interest from academia and industry. Given two HIN nodes$s$and$t$, and a natural number$k$, we study the discovery of the$k$most important meta paths in real time, which can be used to support friend search, product recommendation, anomaly detection, and graph clustering. In this work, we argue that the shortest path between$s$and$t$may not necessarily be the most important path. As such, we combine several ranking functions, which are based onfrequencyandrarity, to redefine the unified importance function of the meta paths between$s$and$t$. Although this importance function can capture more information, it is very time-consuming to find top-$k$meta paths using this importance function. Therefore, we integrate this importance function into a multi-step framework, which can efficiently filter some impossible meta paths between$s$and$t$. In addition, we combine bidirectional searching algorithm with this framework to further boost the efficiency performance. The experiment on different datasets shows that our proposed method outperforms state-of-the-art algorithms in terms of effectiveness with reasonable response time.

Interactive tools make data analysis more efficient and more accessible to end-users by hiding the underlying query complexity and exposing interactive widgets for the parts of the query that matter to the analysis. However, creating custom tailored (i.e., precise) interfaces is very costly, and automated approaches are desirable. We propose a syntactic approach that uses queries from an analysis to generate a tailored interface. We model interface widgets as functions I(q) - > q' that modify the current analysis query q, and interfaces as the set of queries that its widgets can express. Our system, Precision Interfaces, analyzes structural changes between input queries from an analysis, and generates an output interface with widgets to express those changes. Our experiments on the Sloan Digital Sky Survey query log suggest that Precision Interfaces can generate useful interfaces for simple unanticipated tasks, and our optimizations can generate interfaces from logs of up to 10,000 queries in >10s.

Heterogeneous information networks (HINs), which are typed graphs with labeled nodes and edges, have attracted tremendous interest from academia and industry. Given two HIN nodes s and t, and a natural number k, we study the discovery of the k most important paths in real time. The paths found can be used to support friend search, product recommendation, anomaly detection, and graph clustering. Although related algorithms have been proposed before, they were primarily designed to return the k shortest paths from unlabeled graphs. This leads to two problems: (1) there are often many shortest paths between s and t, and so it is not easy to choose the k best ones; and (2) it is arguable whether a shorter path implies a more crucial one. To address these issues, we study the top-k meta path query for a HIN. A meta path abstracts multiple path instances into a high-level path pattern, thereby giving more insight between two nodes. We further study several ranking functions that evaluate the importance of meta paths based on frequency and rarity, rather than on path length. We propose a solution that seamlessly integrates these functions into an A* search framework. The connectivity experiment on ACM dataset shows that our proposed method outperforms state-of-the-art algorithms.

Building interactive tools to support data analysis is hard because it is not always clear what to build and how to build it. To address this problem, we present Precision Interfaces, a semi-automatic system to generate task-specific data analytics interfaces. Precision Interface can turn a log of executed programs into an interface, by identifying micro-variations between the programs and mapping them to interface components. This paper focuses on SQL query logs, but we can generalize the approach to other languages. Our system operates in two steps: it first builds an interaction graph, which describes how the queries can be transformed into each other. Then, it finds a set of UI components that covers a maximal number of transformations. To restrict the domain of changes to be detected, our system uses a domain-specific language, PILang. We describe each of Precision Interface's components, showcase an early prototype on real program logs, and discuss future research opportunities.