HiNet: Hierarchical Classification with Neural Network
This addresses the challenge of hierarchical classification for domains with massive label hierarchies, offering a scalable solution, though it appears incremental as it builds on neural networks for an existing bottleneck.
The paper tackles the problem of classifying large hierarchical labels with over 10,000 distinct traces, which traditional methods like flattening labels or hierarchical SVMs struggle with due to complexity. It introduces a neural network-based hierarchical architecture that is simple to train, with an efficient MAP inference algorithm and reduced complexity from O(n^h) to O(n^2).
Traditionally, classifying large hierarchical labels with more than 10000 distinct traces can only be achieved with flatten labels. Although flatten labels is feasible, it misses the hierarchical information in the labels. Hierarchical models like HSVM by \cite{vural2004hierarchical} becomes impossible to train because of the sheer number of SVMs in the whole architecture. We developed a hierarchical architecture based on neural networks that is simple to train. Also, we derived an inference algorithm that can efficiently infer the MAP (maximum a posteriori) trace guaranteed by our theorems. Furthermore, the complexity of the model is only $O(n^2)$ compared to $O(n^h)$ in a flatten model, where $h$ is the height of the hierarchy.