Stacked Regressions

By Leo Breiman. 1996

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

An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning (2019, methodology): “uite unstable [1], and often results in high-variance or low-condence predictions. An intuitive solution is to train an ensemble of models and use the combined prediction which should be more robust [6,29,54]. However, it is not obvious how to obtain and combine multiple base-learners given the fact”
Machine learning and AI in marketing – Connecting computing power to human insights (2020, methodology): “In stacking, a linear combination of individual predictors is used to achieve better accuracy (Breiman, 1996).”
A Quantile-Based g-Computation Approach to Addressing the Effects of Exposure Mixtures (2019, background): “gativity (or non-positivity) constraint from weighted quantile sum regression. Such constraints have proven valuable in certain machine learning algorithms that combine the results of multiple models (30). However, our approach demonstrates that such constraints can result in magnification of confou”
COMBINING FORECASTING PROCEDURES: SOME THEORETICAL RESULTS (2004, background): “~Breiman, 1996b! have been suggested+ There is evidence suggesting that appropriate weighting of the models ~or procedures! can perform very well~in related contexts, see, e+g+, Buckland, Burnham, and Augustin, 1997; Yang, 2001!+ Promising results have also been obtained under the Bayesian…”
Benchmarking state-of-the-art classification algorithms for credit scoring: An update of research (2015, methodology): “Instead, it uses an additional, second-level classifier to regress base the model predictions (as IVs) on the binary response variable (Breiman, 1996b).”
Autostacker: a compositional evolutionary learning system (2018, methodology): “ in AutoML, such as AutoSklearn [23] and TPOT [30] demonstrate success in a variety of datasets. In this work, we present an AutoML architecture called Autostacker. Inspired by the stacking method [3][6] of ensemble learning, Autostacker automatically discovers pipelines made up of one or many model”
Reliable Data Distillation on Graph Convolutional Network (2020, methodology): “Unlike Bagging [9], which trains each network independently, Snap-shot Ensemble saves the model parameters before resetting the learning rate and treats each model replica as a base model.”
Uncertainty Quantification Using Neural Networks for Molecular Property Prediction (2020, methodology): “36 Scalia et al. previously experimented with summing the outputs of an MPNN MVE with a few ensemble based methods, but there remain many unexplored method combinations and aggregation strategies.”
Aggregating regression procedures to improve performance (2004, methodology): “These include a model selection criterion based method by Buckland et al. (1997), cross-validation based ‘stacking’ by Wolpert (1992) and Breiman (1996) (an earlier 1350–7265 # 2004 ISI/BS version is in Stone 1974), a bootstrap based method by LeBlanc and Tibshirani (1996), a stochastic…”