Neurally-Guided Structure Inference
This addresses the trade-off between speed and generalization in structure inference for tasks like data analysis and parsing, though it appears incremental as it combines existing approaches.
The paper tackled the problem of structure inference by proposing a hybrid algorithm, Neurally-Guided Structure Inference (NG-SI), which uses a neural network to guide hierarchical search, and it outperformed data-driven and search-based alternatives on tasks like probabilistic matrix decomposition and symbolic program parsing.
Most structure inference methods either rely on exhaustive search or are purely data-driven. Exhaustive search robustly infers the structure of arbitrarily complex data, but it is slow. Data-driven methods allow efficient inference, but do not generalize when test data have more complex structures than training data. In this paper, we propose a hybrid inference algorithm, the Neurally-Guided Structure Inference (NG-SI), keeping the advantages of both search-based and data-driven methods. The key idea of NG-SI is to use a neural network to guide the hierarchical, layer-wise search over the compositional space of structures. We evaluate our algorithm on two representative structure inference tasks: probabilistic matrix decomposition and symbolic program parsing. It outperforms data-driven and search-based alternatives on both tasks.