Robust clustering of data collected via crowdsourcing

Abstract

Crowdsourcing approaches rely on the collection of multiple individuals to solve problems that require analysis of large data sets in a timely accurate manner. The inexperience of participants or annotators motivates well robust techniques. Focusing on clustering setups, the data provided by all annotators is suitably modeled here as a mixture of Gaussian components plus a uniformly distributed random variable to capture outliers. The proposed algorithm is based on the expectation-maximization algorithm and allows for soft assignments of data to clusters, to rate annotators according to their performance, and to estimate the number of Gaussian components in the non-Gaussian/Gaussian mixture model, in a jointly manner.

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

“…The main contributions of this paper are the following. Firstly, an unsupervised algorithm for the clustering stage is presented that is similar to our previous work in [20] albeit updated to deal with real data from the MalariaSpot project. Secondly, the complete procedure of clustering and detection, taking into account annotators' performance, is presented using a harmonized notation, which gracefully enables information from the clustering to the detection stage to be conveyed.…”

“…Build Y(s) using(20) in Section 4 and compute {β 0 m (s) : m = 1, • • •,M c (s)}. Compute {β d m (s) : m = 1, • • •,M c (s)} using Alg.…”