paper-with-me

Papers

SparseMAP: Differentiable Sparse Structured Inference

2018-02-12 · ICML 2018 7 · Vlad Niculae, André F. T. Martins, Mathieu Blondel, Claire Cardie

Structured prediction requires searching over a combinatorial number of structures. To tackle it, we introduce SparseMAP: a new method for sparse structured inference, and its natural loss function. SparseMAP automatically selects only a few global structures: it is situated between MAP inference, which picks a single structure, and marginal inference, which assigns probability mass to all structures, including implausible ones. Importantly, SparseMAP can be computed using only calls to a MAP oracle, making it applicable to problems with intractable marginal inference, e.g., linear assignment. Sparsity makes gradient backpropagation efficient regardless of the structure, enabling us to augment deep neural networks with generic and sparse structured hidden layers. Experiments in dependency parsing and natural language inference reveal competitive accuracy, improved interpretability, and the ability to capture natural language ambiguities, which is attractive for pipeline systems.

📄 PDF Abstract BibTeX arXiv:1802.04223

Code (3)

vene/sparsemap 공식 구현 pytorch
mblondel/fenchel-young-losses pytorch
mblondel/projection-losses

Tasks

Dependency ParsingNatural Language InferenceStructured Prediction

Similar Papers 제목 키워드 기반

LP-SparseMAP: Differentiable Relaxed Optimization for Sparse Structured Prediction

2020-01-13 · ICML 2020 1 · Vlad Niculae, André F. T. Martins

Structured prediction requires manipulating a large number of combinatorial structures, e.g., dependency trees or alignments, either as latent or output variables. Recently, the SparseMAP method has been proposed as a di…

Structured Prediction

Sparse and Structured Hopfield Networks

2024-02-21 · Saul Santos, Vlad Niculae, Daniel McNamee, Andre F. T. Martins

Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young lo…

Multiple Instance LearningRetrieval

SparseMap: A Sparse Tensor Accelerator Framework Based on Evolution Strategy

2025-08-18 · Boran Zhao, Haiming Zhai, Zihang Yuan, Hetian Liu 외 arxiv

The growing demand for sparse tensor algebra (SpTA) in machine learning and big data has driven the development of various sparse tensor accelerators. However, most existing manually designed accelerators are limited to …

Reinforcement Learning

Towards Dynamic Computation Graphs via Sparse Latent Structure

2018-09-03 · EMNLP 2018 10 · Vlad Niculae, André F. T. Martins, Claire Cardie

Deep NLP models benefit from underlying structures in the data---e.g., parse trees---typically extracted using off-the-shelf parsers. Recent attempts to jointly learn the latent structure encounter a tradeoff: either mak…

graph construction

Hopfield-Fenchel-Young Networks: A Unified Framework for Associative Memory Retrieval

2024-11-13 · Saul Santos, Vlad Niculae, Daniel McNamee, André F. T. Martins

Associative memory models, such as Hopfield networks and their modern variants, have garnered renewed interest due to advancements in memory capacity and connections with self-attention in transformers. In this work, we …

Image RetrievalMultiple Instance LearningRetrieval