From Wikipedia, the free encyclopedia
This article has multiple issues. Please help
or discuss these issues on the .
This article provides insufficient context for those unfamiliar with the subject. Please help
with a . (January 2013)
This article may be too
for most readers to understand. Please help
this article to , without removing the technical details. The
may contain suggestions. (June 2012)
Conditional random fields (CRFs) are a class of
often applied in
and , where they are used for . Whereas an ordinary
predicts a label for a single sample without regard to "neighboring" samples, a CRF can take
e.g., the linear chain CRF popular in
predicts sequences of labels for sequences of input samples.
CRFs are a type of
. It is used to encode known relationships between observations and construct consistent interpretations. It is often used for
of sequential data, such as natural language text or
and in . Specifically, CRFs find applications in ,
and , among other tasks, being an alternative to the related . In computer vision, CRFs are often used for object recognition and image segmentation.
and Pereira define a CRF on observations
as follows:
be a graph such that
is indexed by the vertices of . Then
is a conditional random field when the random variables , conditioned on , obey the
with respect to the graph: , where
means that
are neighbors in .
What this means is that a CRF is an
whose nodes can be divided into exactly two disjoint sets
and , the observed and output variables, the conditional distribution
is then modeled.
For general graphs, the problem of
in CRFs is intractable. The inference problem for a CRF is basically the same as for an
and the same arguments hold. However there exist special cases for which exact inference is feasible:
If the graph is a chain or a tree, message passing algorithms yield exact solutions. The algorithms used in these cases are analogous to the
for the case of HMMs.
If the CRF only contains pair-wise potentials and the energy is submodular, combinatorial min cut/max flow algorithms yield exact solutions.
If exact inference is impossible, several algorithms can be used to obtain approximate solutions. These include:
Loopy belief propagation
Alpha expansion
Mean field inference
Linear programming relaxations
Learning the parameters
is usually done by
learning for . If all nodes have exponential family distributions and all nodes are observed during training, this
is convex. It can be solved for example using
algorithms, or
such as the
algorithm. On the other hand, if some variables are unobserved, the inference problem has to be solved for these variables. Exact inference is intractable in general graphs, so approximations have to be used.
In sequence modeling, the graph of interest is usually a chain graph. An input sequence of observed variables
represents a sequence of observations and
represents a hidden (or unknown) state variable that needs to be inferred given the observations. The
are structured to form a chain, with an edge between each
and . As well as having a simple interpretation of the
as "labels" for each element in the input sequence, this layout admits efficient algorithms for:
model training, learning the conditional distributions between the
and feature functions from some corpus of training data.
decoding, determining the probability of a given label sequence
inference, determining the most likely label sequence
The conditional dependency of each
is defined through a fixed set of feature functions of the form , which can informally be thought of as measurements on the input sequence that partially determine the
of each possible value for . The model assigns each feature a numerical weight and combines them to determine the probability of a certain value for .
Linear-chain CRFs have many of the same applications as conceptually simpler hidden Markov models (HMMs), but relax certain assumptions about the input and output sequence distributions. An HMM can loosely be understood as a CRF with very specific feature functions that use constant probabilities to model state transitions and emissions. Conversely, a CRF can loosely be understood as a generalization of an HMM that makes the constant transition probabilities into arbitrary functions that vary across the positions in the sequence of hidden states, depending on the input sequence.
Notably in contrast to HMMs, CRFs can contain any number of feature functions, the feature functions can inspect the entire input sequence
at any point during inference, and the range of the feature functions need not have a probabilistic interpretation.
CRFs can be extended into higher order models by making each
dependent on a fixed number
of previous variables . Training and inference are only practical for small values of
(such as o ≤ 5),[] since their computational cost increases exponentially with . Large-margin models for , such as the
can be seen as an alternative training procedure to CRFs.
There exists another generalization of CRFs, the semi-Markov conditional random field (semi-CRF), which models variable-length segmentations of the label sequence . This provides much of the power of higher-order CRFs to model long-range dependencies of the , at a reasonable computational cost.
Latent-dynamic conditional random fields (LDCRF) or discriminative probabilistic latent variable models (DPLVM) are a type of CRFs for sequence tagging tasks. They are
that are trained discriminatively.
In an LDCRF, like in any sequence tagging task, given a sequence of observations x = x?, … x?, the main problem the model must solve is how to assign a sequence of labels y = y?, … y? from one finite set of labels Y. Instead of directly modeling P(y|x) as an ordinary linear-chain CRF would do, instead a set of latent variables h is "inserted" between x and y using the :
This allows the capturing of latent structure between the observations and labels. While LDCRFs can be trained using quasi-Newton methods, a specialized version of the
algorithm called the latent-variable perceptron has been developed for them as well, based on Collins'
algorithm. These models find applications in , specifically
from video streams, and .
This is a partial list of software that implement generic CRF tools.
CRFs based on recurrent neural networks (, )
Linear-chain CRFs with fast online ADF training (, )
Linear-chain CRFs (, )
CRFs with submodular energy functions (, )
General CRFs ()
General CRFs ()
General CRFs ()
Linear-chain CRFs ()
Hidden-state CRFs (, )
Fast linear-chain CRFs ()
Fast restricted linear-chain CRFs ()
Linear-chain CRFs ()
First-order and second-order Markov CRFs ()
First-order, linear-chain CRFs ()
CRF for segmenting images and image volumes ()
This is a partial list of software that implement CRF related tools.
CRF based gene predictor ()
Named Entity Recognizer ()
Named Entity Recognizer ()
Lafferty, J., McCallum, A., Pereira, F. (2001). . "Proc. 18th International Conf. on Machine Learning". Morgan Kaufmann. pp. 282–289.
; Zemel, R.S.; Carreira-Perpin?án, M.A. (2004). "Multiscale conditional random fields for image labeling". IEEE Computer Society. : .
Sha, F., Pereira, F. (2003). .
Settles, B. (2004). . "Proceedings of the International Joint Workshop on Natural Language Processing in Biomedicine and its Applications". pp. 104–107.
Sutton, C McCallum, Andrew (2010). "An Introduction to Conditional Random Fields". : [].
Sarawagi, S William W. Cohen (2005). "". In Lawrence K. Saul, Yair Weiss, Léon Bottou (eds.). "Advances in Neural Information Processing Systems 17". Cambridge, MA: MIT Press. pp. .
Xu S Takuya M Daisuke O Jun'ichi Tsujii (2009). "Latent Variable Perceptron Algorithm for Structured Classification". IJCAI. pp. .
Morency, L. P.; Quattoni, A.; Darrell, T. (2007). . "2007 IEEE Conference on Computer Vision and Pattern Recognition". p. 1. :.  .
T. Lavergne, O. Cappé and F. Yvon (2010). . Proc. 48th Annual Meeting of the , pp. 504-513.
McCallum, A.: Efficiently inducing features of conditional random fields. In: Proc. 19th Conference on Uncertainty in Artificial Intelligence. (2003)
Wallach, H.M.: . Technical report MS-CIS-04-21, University of Pennsylvania (2004)
Sutton, C., McCallum, A.: An Introduction to Conditional Random Fields for Relational Learning. In "Introduction to Statistical Relational Learning". Edited by
and Ben Taskar. MIT Press. (2006)
Klinger, R., Tomanek, K.: Classical Probabilistic Models and Conditional Random Fields. Algorithm Engineering Report TR07-2-013, Department of Computer Science, Dortmund University of Technology, December 2007. ISSN .
: Hidden categories:conditional response是什么意思及反义词
沪江词库精选conditional response是什么意思、用法及解释、英语短语。
conditional response的中文解释： phr. 条件反应
同义词：，，，，，。
an acquired response that is under the control of (conditional on the occurrence of) a stimulus
英语短语推荐
2014conditional response是什么意思由沪江网提供。conditionals
英 [k?n'd???n?lz]
美 [k?n'd???n?lz]
conditionals的用法和样例：
In the COBOL system, the information represented by a data item, arithmetic expression, or conditional expression.
在COBOL系统中，由一个数据项、算术表达式或条件表达式表示的信息。
conditionals的海词问答与网友补充：
conditionals的相关资料：
相关词典网站：conditional
英[k?n'd???n?l]
美[k?n'd???n?l]
adj. 有条件的；条件的n. 条件句
Their support is conditional on his proposals meeting their approval.
他们的支持是以他的提案得到他们的批准为条件的。
This is a conditional ceasefire.
这是有条件停火。
"If we go" is a conditional clause.
“要是我们去的话”是条件子句。
This is called the antecedent of a conditional statement.
这被称为条件句前提。
1.imposing or depending on or containing a condition
conditional acceptance of the terms
lent conditional support
the conditional sale will not be complete until the full purchase price is paid
2.qualified by reservations
只有登录后，才能查看此项，现在是否？
1.This is a conditional ceasefire.
这是有条件停火。
2.Is there a remedy for this condition?
有没有解决这个情况的方法？
3.The condition of man . . . is a condition of war of everyone against everyone
人类的状态.是人与人对抗的战争状态
4.A rheumatic condition,pain,joint
风湿病、风湿痛、患风湿病的关节.
5.Conditional approval,acceptance,etc
含有条件的批准、接受等