Message Passing for Soft Constraint Dual Decomposition

نویسندگان

  • David Belanger
  • Alexandre Passos
  • Sebastian Riedel
  • Andrew McCallum
چکیده

Dual decomposition provides the opportunity to build complex, yet tractable, structured prediction models using linear constraints to link together submodels that have available MAP inference routines. However, since some constraints might not hold on every single example, such models can often be improved by relaxing the requirement that these constraints always hold, and instead replacing them with soft constraints that merely impose a penalty if violated. A dual objective for the resulting MAP inference problem differs from the hard constraint problem’s associated dual decomposition objective only in that the dual variables are subject to box constraints. This paper introduces a novel primaldual block coordinate descent algorithm for minimizing this general family of box-constrained objectives. Through experiments on two natural language corpus-wide inference tasks, we demonstrate the advantages of our approach over the current alternative, based on copying variables, adding auxiliary submodels and using traditional dual decomposition. Our algorithm performs inference in the same model as was previously published for these tasks, and thus is capable of achieving the same accuracy, but provides a 2-10x speedup over the current state of the art.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Institut für Numerische und Angewandte Mathematik Parallel Total Variation Minimization

In [ROF92] Rudin, Osher and Fatemi introduced a denoising algorithm using total variation regularization. In this work we provide a parallel algorithm for this variational minimization problem. It is based on the primal-dual formulation and hence leads to solve the saddle point problem for the primal and dual variable. For that reason Newton’s method with damping was used. The arising constrain...

متن کامل

Distributed Clustering in Cognitive Radio Ad Hoc Networks Using Soft-Constraint Affinity Propagation

Absence of network infrastructure and heterogeneous spectrum availability in cognitive radio ad hoc networks (CRAHNs) necessitate the self-organization of cognitive radio users (CRs) for efficient spectrum coordination. The cluster-based structure is known to be effective in both guaranteeing system performance and reducing communication overhead in variable network environment. In this paper, ...

متن کامل

Unifying Tree-Decomposition Schemes

The paper provides a unifying perspective of tree-decomposition algorithms appearing in various automated reasoning areas, such as join-tree clustering for constraint-satisfaction and the clique-tree algorithm for probabilistic reasoning. Subsequently, the paper extends the variable-elimination scheme called bucket-elimination (BE) into a two-phase message passing along a bucket-tree (BTE), mak...

متن کامل

Local Consistency in Junction Graphs for Constraint-Based Inference

The concept of local consistency plays a central role in constraint satisfaction. Given a constraint satisfaction problem (CSP), local consistency can be characterized as deriving new, possibly tighter, constraints based on local information. The derived constraints simplify the representation of the original CSP without the loss of solutions. This can be seen as a preprocessing procedure. Base...

متن کامل

Adaptive Domain Abstraction in a Soft-Constraint Message-Passing Algorithm

The computational tasks of model-based diagnosis and planning in embedded systems can be framed as soft-constraint optimization problems with planning costs or state transition probabilities as preferences. Running constraint optimization in embedded systems requires to reduce complexity, which can be achieved by combining dynamic programming message-passing algorithms with message approximatio...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2014