Search papers, labs, and topics across Lattice.
The paper introduces tensorFM, a novel model for tabular categorical data that efficiently captures high-order feature interactions using a low-rank tensor approximation. This approach addresses the challenge of modeling complex relationships between categorical attributes (fields) in applications like click-through rate prediction. The authors demonstrate that tensorFM achieves competitive performance compared to state-of-the-art methods while maintaining low latency, making it suitable for real-time applications.
Achieve state-of-the-art performance on tabular data with a low-latency model that efficiently captures high-order feature interactions via a low-rank tensor approximation.
We address prediction problems on tabular categorical data, where each instance is defined by multiple categorical attributes, each taking values from a finite set. These attributes are often referred to as fields, and their categorical values as features. Such problems frequently arise in practical applications, including click-through rate prediction and social sciences. We introduce and analyze {tensorFM}, a new model that efficiently captures high-order interactions between attributes via a low-rank tensor approximation representing the strength of these interactions. Our model generalizes field-weighted factorization machines. Empirically, tensorFM demonstrates competitive performance with state-of-the-art methods. Additionally, its low latency makes it well-suited for time-sensitive applications, such as online advertising.
