Unsupervised online clustering and detection algorithms using crowdsourced data for malaria diagnosis

Pagès-Zamora, Alba; Cabrera-Bean, Margarita; Diaz-Vilor, Carles

doi:10.1016/j.patcog.2018.09.001

Cited by 13 publications

(8 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Today, after a little more than 40 years, it is still one of the most popular algorithms for statistical pattern recognition. Studies range from the theoretical, i.e., convergence of the EM and its variant DA-EM in [2], to modifications of the EM algorithm for different purposes: image matching [3]; parameter estimation [4,5]; malaria diagnoses [6]; mixture simplification [7]; and audio-visual scene analysis [8].…”

Section: Introductionmentioning

confidence: 99%

“…Once estimated, the parameters of MM can be used multi-purposely. Their use is equally interesting for density-estimation tasks [7,11] and clustering tasks [6,12,13]. In the context of MM parameter estimation, the EM algorithm can be seen as a clustering algorithm that maximizes the missing data log-likelihood function as the objective function.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Improved Initialization of the EM Algorithm for Mixture Model Parameter Estimation

2020

View full text Add to dashboard Cite

A commonly used tool for estimating the parameters of a mixture model is the Expectation–Maximization (EM) algorithm, which is an iterative procedure that can serve as a maximum-likelihood estimator. The EM algorithm has well-documented drawbacks, such as the need for good initial values and the possibility of being trapped in local optima. Nevertheless, because of its appealing properties, EM plays an important role in estimating the parameters of mixture models. To overcome these initialization problems with EM, in this paper, we propose the Rough-Enhanced-Bayes mixture estimation (REBMIX) algorithm as a more effective initialization algorithm. Three different strategies are derived for dealing with the unknown number of components in the mixture model. These strategies are thoroughly tested on artificial datasets, density–estimation datasets and image–segmentation problems and compared with state-of-the-art initialization methods for the EM. Our proposal shows promising results in terms of clustering and density-estimation performance as well as in terms of computational efficiency. All the improvements are implemented in the rebmix R package.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Improved Initialization of the EM Algorithm for Mixture Model Parameter Estimation

2020

View full text Add to dashboard Cite

show abstract

“…EM's popularity has risen due to its use in estimating mixture-model parameters [23,24]. The use of estimated mixture-models is equally interesting for densityestimation tasks [20,25] and clustering tasks [26][27][28]. For clustering, EM aims at finding clusters such that maximum likelihood of each cluster's parameters are obtained.…”

Section: Study Logic and Technology Pathway 211 Study Logicmentioning

confidence: 99%

Bicycle Speed Modelling Considering Cyclist Characteristics, Vehicle Type and Track Attributes

Yan

Chen

et al. 2021

WEVJ

View full text Add to dashboard Cite

Cycling is an increasingly popular mode of transport as part of the response to air pollution, urban congestion, and public health issues. The emergence of bike sharing programs and electric bicycles have also brought about notable changes in cycling characteristics, especially cycling speed. In order to provide a better basis for bicycle-related traffic simulations and theoretical derivations, the study aimed to seek the best distribution for bicycle riding speed considering cyclist characteristics, vehicle type, and track attributes. K-means clustering was performed on speed subcategories while selecting the optimal number of clustering using L method. Then, 15 common models were fitted to the grouped speed data and Kolmogorov–Smirnov test, Akaike information criterion, and Bayesian information criterion were applied to determine the best-fit distribution. The following results were acquired: (1) bicycle speed sub-clusters generated by the combinations of bicycle type, bicycle lateral position, gender, age, and lane width were grouped into three clusters; (2) Among the common distribution, generalized extreme value, gamma and lognormal were the top three models to fit the three clusters of speed dataset; and (3) integrating stability and overall performance, the generalized extreme value was the best-fit distribution of bicycle speed.

show abstract

“…Applications are found in diverse areas such as medicine and biology, e.g. [2] [3] where decision agents are either algorithmic techniques or individuals, respectively; team decisionmaking strategies [4]; and 5G communication systems [5], where decision agents are sensors.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised Ensemble Classification With Correlated Decision Agents

Cabrera-Bean

Pagès-Zamora

Diaz-Vilor

2019

IEEE Signal Process. Lett.

Self Cite

View full text Add to dashboard Cite

Decision-making procedures when a set of individual binary labels is processed to produce a unique joint decision can be approached modeling the individual labels as multivariate independent Bernoulli random variables. This probabilistic model allows an unsupervised solution using EM-based algorithms, which basically estimate the distribution model parameters and take a joint decision using a Maximum a Posteriori criterion. These methods usually assume that individual decision agents are conditionally independent, an assumption that might not hold in practical setups. Therefore, in this work we formulate and solve the decision-making problem using an EM-based approach but assuming correlated decision agents. Improved performance is obtained on synthetic and real datasets, compared to classical and state-of-the-art algorithms.

show abstract

Unsupervised online clustering and detection algorithms using crowdsourced data for malaria diagnosis

Cited by 13 publications

References 21 publications

Improved Initialization of the EM Algorithm for Mixture Model Parameter Estimation

Improved Initialization of the EM Algorithm for Mixture Model Parameter Estimation

Bicycle Speed Modelling Considering Cyclist Characteristics, Vehicle Type and Track Attributes

Unsupervised Ensemble Classification With Correlated Decision Agents

Contact Info

Product

Resources

About