NASH: Toward End-to-End Neural Architecture for Generative Semantic Hashing
This work addresses the need for more efficient and theoretically grounded semantic hashing methods in information retrieval systems, representing an incremental advancement by integrating variational inference and rate-distortion theory into an end-to-end framework.
The paper tackles the problem of two-stage training and ad-hoc binary constraint handling in semantic hashing by proposing NASH, an end-to-end neural architecture that treats binary hashing codes as Bernoulli latent variables and uses neural variational inference for training, resulting in significant performance improvements over state-of-the-art models on three public datasets.
Semantic hashing has become a powerful paradigm for fast similarity search in many information retrieval systems. While fairly successful, previous techniques generally require two-stage training, and the binary constraints are handled ad-hoc. In this paper, we present an end-to-end Neural Architecture for Semantic Hashing (NASH), where the binary hashing codes are treated as Bernoulli latent variables. A neural variational inference framework is proposed for training, where gradients are directly back-propagated through the discrete latent variable to optimize the hash function. We also draw connections between proposed method and rate-distortion theory, which provides a theoretical foundation for the effectiveness of the proposed framework. Experimental results on three public datasets demonstrate that our method significantly outperforms several state-of-the-art models on both unsupervised and supervised scenarios.