Figure 1
From: Bayesian clustering and feature selection for cancer tissue samples
Comparison of running times of the algorithms. Results of the comparison of the running times of alternative algorithms (MCMC Gibbs sampling, greedy algorithm) on artificial data sets using the basic beta-binomial model. The x-axis shows the complexity level of the data set in terms of the number of clusters and the y-axis shows the average time for five data sets at each level in seconds. Because the times taken by the two versions of the greedy algorithm were roughly equal, they are represented by just one curve (red asterisks). The MCMC initialized by a single cluster is represented by a green square. The MCMC initialized with the correct number plus additional 10 clusters is represented by a blue circle. The times recorded for the MCMC correspond to the times when they reached a model with equal or higher marginal likelihood than the highest value found by the greedy algorithm. If the MCMC never found such a model, the recorded time is the time taken by the MCMC to reach the model with the highest marginal likelihood in that run.