Metalearning
A Metalearning is a Machine Learning Algorithm that is applied on metadata about machine learning algorithms.
- AKA: Adaptive Learning.
- Example(s):
- …
- Counter-Example(s):
- See: Metaheuristic, Machine Learning, Meta-Data, Learning Algorithms, Inductive Bias, Data Mining, Database, Inductive Transfer.
References
2018
- (Wikipedia, 2018) ⇒ https://en.wikipedia.org/wiki/Meta_learning_(computer_science) Retrieved:2018-4-8.
- Meta learning is a subfield of machine learning where automatic learning algorithms are applied on metadata about machine learning experiments. As of 2017 the term had not found a standard interpretation, however the main goal is to use such metadata to understand how automatic learning can become flexible in solving learning problems, hence to improve the performance of existing learning algorithms or to learn (induce) the learning algorithm itself, hence the alternative term learning to learn.
Flexibility is important because each learning algorithm is based on a set of assumptions about the data, its inductive bias. This means that it will only learn well if the bias matches the learning problem. A learning algorithm may perform very well in one domain, but not on the next. This poses strong restrictions on the use of machine learning or data mining techniques, since the relationship between the learning problem (often some kind of database) and the effectiveness of different learning algorithms is not yet understood.
By using different kinds of metadata, like properties of the learning problem, algorithm properties (like performance measures), or patterns previously derived from the data, it is possible to learn, select, alter or combine different learning algorithms to effectively solve a given learning problem. Critiques of meta learning approaches bear a strong resemblance to the critique of metaheuristic, a possibly related problem. A good analogy to meta-learning, and the inspiration for Bengio et al.'s early work (1991), considers that genetic evolution learns the learning procedure encoded in genes and executed in each individual's brain.
- Meta learning is a subfield of machine learning where automatic learning algorithms are applied on metadata about machine learning experiments. As of 2017 the term had not found a standard interpretation, however the main goal is to use such metadata to understand how automatic learning can become flexible in solving learning problems, hence to improve the performance of existing learning algorithms or to learn (induce) the learning algorithm itself, hence the alternative term learning to learn.
2017
- (Brazdil et al. 2017) ⇒ Brazdil P., Vilalta R., Giraud-Carrier C., Soares C. (2017) Metalearning. In: Sammut, C., Webb, G.I. (eds) Encyclopedia of Machine Learning and Data Mining. Springer, Boston, MA
- ABSTRACT: In the area machine learning / data mining many diverse algorithms are available nowadays and hence the selection of the most suitable algorithm may be a challenge. This is aggravated by the fact that many algorithms require that certain parameters be set. If a wrong algorithm and/or parameter configuration is selected, substandard results may be obtained. The topic of metalearning aims to facilitate this task. Metalearning typically proceeds in two phases. First, a given set of algorithms A (e.g. classification algorithms) and datasets D is identified and different pairs < ai,dj > from these two sets are chosen for testing. The dataset di is described by certain meta-features which together with the performance result of algorithm ai constitute a part of the metadata. In the second phase the metadata is used to construct a model, usually again with recourse to machine learning methods. The model represents a generalization of various base-level experiments. The model can then be applied to the new dataset to recommend the most suitable algorithm or a ranking ordered by relative performance. This article provides more details about this area. Besides, it discusses also how the method can be combined with hyperparameter optimization and extended to sequences of operations (workflows).