Model-Based Clustering Task

A Model-Based Clustering Task is a clustering task that produces a model-based learning task.

• Context:
  • It can be solved by Model-Based Clustering System by implementing Model-Based Clustering Algorithms.
• Example(s):
  • …
• Counter-Example(s):
  • a Model-Free Clustering Task.
  • a Model-based Supervised Learning Task, such as DTree learning, or HMM learning.
  • Graphical Model Training Task,
  • Probabilistic Model Training Task,
• See: Model-Based Reinforcement Learning, Model-Based Control, Generative Model, Mixture Model, Statistical Modelling, Machine Learning, Data Mining, Mathematical Modeling.

---

---

References

2017

• (Banerjee & Shan, 2017) ⇒ Arindam Banerjee, and Hanhuai Shan (2017). "Model-Based Clustering" In: (Sammut & Webb, 2017). DOI:10.1007/978-1-4899-7687-1_554
  • QUOTE: Model-based clustering is a statistical approach to data clustering. The observed (multivariate) data is assumed to have been generated from a finite mixture of component models. Each component model is a probability distribution, typically a parametric multivariate distribution. For example, in a multivariate Gaussian mixture model, each component is a multivariate Gaussian distribution. The component responsible for generating a particular observation determines the cluster to which the observation belongs. However, the component generating each observation as well as the parameters for each of the component distributions are unknown. The key learning task is to determine the component responsible for generating each observation, which in turn gives the clustering of the data. Ideally, observations generated from the same component are inferred to belong to the same cluster. In addition to inferring the component assignment of observations, most popular learning approaches also estimate the parameters of each component in the process. The strength and popularity of the methods stem from the fact that they are applicable for a wide variety of data types, such as multivariate, categorical, sequential, etc., as long as suitable component generative models can be constructed. Such methods have found applications in several domains such as text clustering, image processing, computational biology, and climate sciences.

---