Using Meta-Learning to Support Data Mining

By Ricardo Vilalta, Christophe G. Giraud-Carrier, Pavel Brazdil, Carlos Soares. 2004

393 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

A Survey on Neural Architecture Search (2019, methodology): “Most works that examine both search spaces, the global and the cell-based space, support this choice by providing empirical evidence that well-performing can be obtained in the cell-based search space (Pham et al., 2018). Regardless, cell-based search spaces benefit from the advantage that discovere”
Initializing Bayesian Hyperparameter Optimization via Meta-Learning (2015, methodology): “• Landmarking metafeatures (Pfahringer, Bensusan, and Giraud-Carrier 2000) are computed by running several fast machine learning algorithms on the dataset.”
Meta-Learning in Neural Networks: A Survey (2020, neutral): “In contrast, previous surveys are rather out of date in this fast moving field, and/or focus on algorithm selection for data mining [29], [31], [33]–[37], AutoML [32],”
Auto-sklearn: Efficient and Robust Automated Machine Learning (2019, background): “Notably, we had to exclude the prominent and eﬀective category of landmarking meta-features [31] (which measure the performance of simple base learners), because they were computationally too expensive to be helpful in the online evaluation phase.”
Meta-learning for evolutionary parameter optimization of classifiers (2012, background): “Statistical meta-features (Gama and Brazdil 1995; Vilalta et al. 2004; Brazdil et al. 1994; King et al. 1995): – cancor1: canonical correlation for the best single combination of features – cancor2: canonical correlation for the best single combination of features orthogonal to cancor1 – fract1:…”
Fair lending needs explainable models for responsible recommendation (2018, background): “Examples of the latter include being in the highest risk twentile predicted by another model, or the largest principal component built from all credit bureau information, or even a landmark feature produced by running a fast (if less accurate) machine learning algorithm [42].”
Automatic classifier selection for non-experts (2014, methodology): “The landmarking approach utilizes simple and fast computable classification algorithms [27, 30, 6].”
Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms (2012, unknown): “Categories and Subject Descriptors I.2.6 [Artificial Intelligence]: Learning; I.2.2 [Artificial Intelligence]: Automatic Programming; G.1.6 [Mathematics of Computing]: Optimization General Terms Algorithms, Performance, Experimentation Keywords Model selection; Hyperparameter optimization; WEKA”
Cross-disciplinary perspectives on meta-learning for algorithm selection (2009, methodology): “The idea of landmarking as an alternative to generating computationally expensive features was also developed further during METAL to create models predicting the performance of algorithms relative to the performance of simpler and faster learners [Pfahringer et al. 2000; Furnkranz and Petrak 2001; ”
Beyond Manual Tuning of Hyperparameters (2015, methodology): “such as LDA, Naive Bayes and decision trees on the target dataset/task have been used as meta-features, called landmark features [50].”
Machine Learning - The Art and Science of Algorithms that Make Sense of Data (2012, unknown): “The idea of training simple models to obtain further data characteristics is known as landmarking (Pfahringer et al., 2000).”
Does Feature Selection Improve Classification? A Large Scale Experiment in OpenML (2016, background): “A classifier can then learn to predict for new data sets which algorithm will perform best [22].”
On the predictive power of meta-features in OpenML (2017, methodology): “Landmarking measures have been evaluated and have shown to perform well in many works (Fürnkranz and Petrak, 2001; Bensusan and Kalousis, 2001; Pfahringer et al., 2000; Reif et al., 2014).”
A Perspective View and Survey of Meta-Learning (2002, background): “From this point of view, meta-learning is the process of finding areas of expertise of learners called landmarkers, and of correlating those areas with the performance of other – more advanced – learners (Bensusan and Giraud-Carrier (2000), Pfahringer et al. (2000)).”
Combination of Task Description Strategies and Case Base Properties for Meta-Learning (2002, methodology): “A completely different approach was chosen by [15] and [2].”