A comprehensive review of conditional random fields: variants, hybrids and applications

Yu, Bengong; Fan, Zhaodi

doi:10.1007/s10462-019-09793-6

Cited by 17 publications

(14 citation statements)

References 153 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…9 CRFs specifically have played a crucial role in sequential modeling tasks and have been used extensively in areas such as natural language processing (NLP), where they frequently outperform their generative counterparts. 14,15 Recently, deep learning approaches have become popular methods for processing sequential data. However, these models often require a great deal of training data and/or pretraining efforts to show marked improvements over classical ML models.…”

Section: Discussionmentioning

confidence: 99%

“…1 and Supplementary Figure S1) have been shown to outperform generative models, such as HMMs, in various application domains. 14,15 Furthermore, compared to their "black box" RNN counterparts, CRFs have the advantage of being inherently interpretable, an important feature in a biomedical context. 16 Here, we describe GECCO (GEne Cluster prediction with COnditional random fields; https://gecco.embl.de), a high-precision, scalable method for de novo BGC identification in microbial genomic and metagenomic data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Accuratede novoidentification of biosynthetic gene clusters with GECCO

Lm¹,

Larralde²,

Fleck

et al. 2021

Preprint

View full text Add to dashboard Cite

Biosynthetic gene clusters (BGCs) are enticing targets for (meta)genomic mining efforts, as they may encode novel, specialized metabolites with potential uses in medicine and biotechnology. Here, we describe GECCO (GEne Cluster prediction with COnditional random fields; https://gecco.embl.de), a high-precision, scalable method for identifying novel BGCs in (meta)genomic data using conditional random fields (CRFs). Based on an extensive evaluation of de novo BGC prediction, we found GECCO to be more accurate and over 3x faster than a state-of-the-art deep learning approach. When applied to over 12,000 genomes, GECCO identified nearly twice as many BGCs compared to a rule-based approach, while achieving higher accuracy than other machine learning approaches. Introspection of the GECCO CRF revealed that its predictions rely on protein domains with both known and novel associations to secondary metabolism. The method developed here represents a scalable, interpretable machine learning approach, which can identify BGCs de novo with high precision.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Accuratede novoidentification of biosynthetic gene clusters with GECCO

Lm¹,

Larralde²,

Fleck

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Conditional Random Fields (CRFs), as an important and prevalent type of machine learning method, are designed for building probabilistic models to segment and label sequence data. The CRF is an undirected discriminative graphical model focusing on posterior distribution of observation and possible label sequence as opposed to the generative nature of MRF [8]. The CRFs are developed on the basis of Maximum Entropy Markov Models (MEMMs) in 2002 [9], with the aim at avoiding the fundamental limitations of it and other directed graphical models like Hidden Markov Model (HMM) [10].…”

Section: B Conditional Random Fieldsmentioning

confidence: 99%

“…However, some existing methods do not take the contextual information on neighboring labels into consideration. For example, some traditional binary classifiers like Support Vector Machine (SVM) and Maximum Entropy only consider one single input and ignore the spatial relationship with other inputs while predicting the labels [8]. Besides, the advanced DL model Convolutional Neural Network (CNN) also has this problem.…”

Section: Motivationmentioning

confidence: 99%

A Comprehensive Review for MRF and CRF Approaches in Pathology Image Analysis

Li,

et al. 2020

Preprint

View full text Add to dashboard Cite

Pathology image analysis is an essential procedure for clinical diagnosis of many diseases. To boost the accuracy and objectivity of detection, nowadays, an increasing number of computer-aided diagnosis (CAD) system is proposed. Among these methods, random field models play an indispensable role in improving the analysis performance. In this review, we present a comprehensive overview of pathology image analysis based on the markov random fields (MRFs) and conditional random fields (CRFs), which are two popular random field models. Firstly, we introduce the background of two random fields and pathology images. Secondly, we summarize the basic mathematical knowledge of MRFs and CRFs from modelling to optimization. Then, a thorough review of the recent research on the MRFs and CRFs of pathology images analysis is presented. Finally, we investigate the popular methodologies in the related works and discuss the method migration among CAD field.

show abstract

“…The conditional random fields (CRF) were engaged as a recurrent neural network to detect prostate cancer [14]. Here CRF is intended to capture a probability distribution from images' observations, as described in some works cited in [15].…”

Section: Introductionmentioning

confidence: 99%

Decomposed Mean Euler-Poincaré Characteristic Model for a Non-Gaussian Physiological Random Field

Ramos-Martinez

Corbier²,

Alvarado

et al. 2021

IEEE Access

View full text Add to dashboard Cite

This paper introduces a new approach of the mean Euler-Poincaré characteristic for non-Gaussian random fields (NGRF), which is based on the decomposition by a basic function named motherwave. The method is proved for long-term recorded, noisy physiological signals. A pretreatment allows the signal to become smooth as the original one is fitted through a Random Algebraic Polynomials (RAP)-based scheme. After that, the polynomized signals are merged by thresholding the RAP function at different levels u. In this way, it is formed a real-valued non-Gaussian physiological random field (NGPRF). Thereby, we deal with their geometric properties centered on their excursion sets A u (Φ, T ) and a topological invariant, such as the Euler Poincaré Characteristic (EPC) ϕ(A u (Φ, T )). The highlight of this work is an explicit model, referred to as the decomposed mean Euler-Poincaré characteristic (DMEPC). The proposed method produces a reduced model with a viable interpretation for different heart conditions investigated for data issued from Holter EKG recordings.

show abstract

A comprehensive review of conditional random fields: variants, hybrids and applications

Cited by 17 publications

References 153 publications

Accuratede novoidentification of biosynthetic gene clusters with GECCO

Accuratede novoidentification of biosynthetic gene clusters with GECCO

A Comprehensive Review for MRF and CRF Approaches in Pathology Image Analysis

Decomposed Mean Euler-Poincaré Characteristic Model for a Non-Gaussian Physiological Random Field

Contact Info

Product

Resources

About