Modeling Concept Drift: A Probabilistic Graphical Model Based Approach

Borchani, Hanen; Martínez, Ana María; Masegosa, Andrés R.; Langseth, Helge; Nielsen, Thomas D.; Salmerón, Antonio; Fernández, Antonio; Madsen, Anders L.; Sáez, Ramón

doi:10.1007/978-3-319-24465-5_7

Cited by 15 publications

(34 citation statements)

References 13 publications

(22 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The output generated by Code Fragment 7 is shown below. The print-out includes the distributions associated to each variable: the discrete variables have a multinomial, while the continuous ones have normal distributions given by (2). Note that, for space restriction, the generated output has been reduced.…”

Section: Static Dynamicmentioning

confidence: 99%

“…In this way, AMIDST's approach to machine learning is based on the use of openbox models that can be inspected and which can incorporate prior information or knowledge about the domain, in contrast to other approaches based on blackbox models, which cannot be interpreted by the users. This is why the focus of AMIDST is not only to learn models for making predictions rather than to learn models to extract knowledge from the data [2,1]. And, as commented above, with the advantage that this toolbox is equipped with a general learning engine implementing scalable variational Bayesian inference algorithms, which allow to learn models exploiting a wide range of computational settings.…”

Section: Introductionmentioning

confidence: 99%

“…The toolbox has been developed by the AMIDST Consortium 1 , which includes both academic and industrial partners. Our software has been used in the automotive industry [20] and in the finance sector [1,2]. The AMIDST Toolbox can be used, for instance, for classification, clustering, regression, and density estimation tasks [16].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

AMIDST: A Java toolbox for scalable probabilistic machine learning

Masegosa

Martínez

Ramos-López

et al. 2019

Knowledge-Based Systems

Self Cite

View full text Add to dashboard Cite

The AMIDST Toolbox is a software for scalable probabilistic machine learning with a special focus on (massive) streaming data. The toolbox supports a flexible modeling language based on probabilistic graphical models with latent variables and temporal dependencies. The specified models can be learnt from large data sets using parallel or distributed implementations of Bayesian learning algorithms for either streaming or batch data. These algorithms are based on a flexible variational message passing scheme, which supports discrete and continuous variables from a wide range of probability distributions. AMIDST also leverages existing functionality and algorithms by interfacing to software tools such as Flink, Spark, MOA, Weka, R and HUGIN. AMIDST is an open source toolbox written in Java and available at http://www.amidsttoolbox.com under the Apache Software License version 2.0.

show abstract

Section: Static Dynamicmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

AMIDST: A Java toolbox for scalable probabilistic machine learning

Masegosa

Martínez

Ramos-López

et al. 2019

Knowledge-Based Systems

Self Cite

View full text Add to dashboard Cite

show abstract

“…It can be a good idea to perform, for instance, a parallel cross-validation evaluation in order to determine its value. We have so far performed experiments using real datasets of different sizes and observed that the optimal batch size is generally in the order of one thousand [49,50].…”

Section: The Spliterator Interfacementioning

confidence: 99%

Probabilistic Graphical Models on Multi-Core CPUs Using Java 8

Masegosa

Martínez

Borchani

2016

IEEE Comput. Intell. Mag.

Self Cite

View full text Add to dashboard Cite

In this paper, we discuss software design issues related to the development of parallel computational intelligence algorithms on multi-core CPUs, using the new Java 8 functional programming features. In particular, we focus on probabilistic graphical models (PGMs) and present the parallelisation of a collection of algorithms that deal with inference and learning of PGMs from data. Namely, maximum likelihood estimation, importance sampling, and greedy search for solving combinatorial optimisation problems. Through these concrete examples, we tackle the problem of defining efficient data structures for PGMs and parallel processing of same-size batches of data sets using Java 8 features. We also provide straightforward techniques to code parallel algorithms that seamlessly exploit multi-core processors. The experimental analysis, carried out using our open source AMIDST (Analysis of MassIve Data STreams) Java toolbox, shows the merits of the proposed solutions. of using several processing cores on a chip, industry was able to multiply the computing performance of the processors while keeping the clock frequency fixed.Computational intelligence methods have been strongly influenced by the emergence of parallel computing architectures. The literature on distributed and parallel algorithms has exploded in the last years [3][4][5][6][7][8]. Consequently, computational intelligence software libraries, especially in the field of "Big data", have emerged to help developers to leverage the computational power of systems with multiple processing units [9,10]. Classic scientific programming libraries like R or Matlab have also tried to adapt to multi-core CPUs or computer clusters by providing specialised libraries [11][12][13].However, programming computational intelligence methods on parallel architectures is not an easy task yet. Apart from the latter developments, there are many other parallel programming languages such as Orca [14], Occam ABCL [15], SNOW [16], MPI [17], PARLOG [18], and Scala [19]. Nevertheless, at the time of writing, none of them seems to be widely accepted by the computational intelligence community. For example, if we take a look at some reference venues such as the JMLR Software track 1 , IEEE Fuzzy Software for Soft Computing series [20], and the Journal of Statistical Software 2 , the vast majority of the proposed software libraries do not rely on any of these parallel programming languages [?]. Moreover, only a few released libraries contain implementations of computational intelligence techniques that are able to exploit multi-core CPUs [21-24]. Enterprisesupported scientific computing libraries constitute an exception [12], because in general programming concurrent and parallel applications is tough, time consuming, error prone, and requires specific programming skills that may be beyond the expertise of C. I. researchers.The latest Java release (JDK 8.0, March 2014) is an attempt to bring a new API to code transparent and seamless parallel applications to the general public (that is, non-speci...

show abstract

“…For that, we use the conditional linear Gaussian (CLG) model, which is a kind of Bayesian network with continuous and discrete nodes. Our model is based on the one proposed by Borchani et al [1,4], an approach using the CLG model and latent variables that was applied to continuous variables. Basically, we propose transforming the discrete data into continuous and then applying a similar approach with latent variables.…”

Section: Introductionmentioning

confidence: 99%

Virtual Subconcept Drift Detection in Discrete Data Using Probabilistic Graphical Models

Cabañas

Cano

Gómez-Olmedo

et al. 2018

Communications in Computer and Information Science

Self Cite

View full text Add to dashboard Cite

A common problem in mining data streams is that the distribution of the data might change over time. This situation, which is known as concept drift, should be detected for ensuring the accuracy of the models. In this paper we propose a method for concept drift detection in discrete streaming data using probabilistic graphical models. In particular, our approach is based on the use of conditional linear Gaussian Bayesian networks with latent variables. We demonstrate and analyse the proposed model using synthetic and real data.

show abstract

Modeling Concept Drift: A Probabilistic Graphical Model Based Approach

Cited by 15 publications

References 13 publications

AMIDST: A Java toolbox for scalable probabilistic machine learning

AMIDST: A Java toolbox for scalable probabilistic machine learning

Probabilistic Graphical Models on Multi-Core CPUs Using Java 8

Virtual Subconcept Drift Detection in Discrete Data Using Probabilistic Graphical Models

Contact Info

Product

Resources

About