Local case-control sampling: Difference between revisions

Revision as of 17:44, 14 June 2015

Revision as of 19:47, 12 August 2015

Line 1:

In [[machine learning]], '''local case-control sampling''' <ref name="LCC">{{cite journal|last1=Fithian|first1=William|last2=Hastie|first2=Trevor|title=Local case-control sampling: Efficient subsampling in imbalanced data sets|journal=The Annals of Statistics|date=2014|volume=42|issue=5|pagepages=1693–1724|ref=http://arxiv.org/abs/1306.3706}}</ref> is an [[algorithm]] used to reduce the complexity of training a [[logistic regression]] classifier. The algorithm reduces the training complexity by selecting a small subsample of the original dataset for training. It assumes the availability of a (unreliable) pilot estimation of the parameters. It then performs a single pass over the entire dataset using the pilot estimation to identify the most "surprising" samples. In practice, the pilot may come from prior knowledge or training using a subsample of the dataset. The algorithm is most effective when the underlying dataset is imbalanced. It exploits the structures of conditional imbalanced datasets more efficiently than alternative methods, such as [[Logistic_regressionLogistic regression#Case-control_samplingcontrol sampling|case control sampling]] and weighted case control sampling.

== Imbalanced datasets ==

Line 15:

# The output model is <math> \hat{\theta} = (\hat{\alpha}, \hat{\beta}) </math>, where <math>\hat{\alpha} \leftarrow \hat{\alpha}_S + \tilde{\alpha} </math> and <math>\hat{\beta} \leftarrow \hat{\beta}_S + \tilde{\beta} </math>.

The algorithm can be understood as selecting samples that surprises the pilot model. Intuitively these samples are closer to the [[Decision boundary|decision boundary]] of the classifier and is thus more informative.

=== Obtaining the pilot model ===

In practice, for cases where a pilot model is naturally available, the algorithm can be applied directly to reduce the complexity of training. In cases where a natural pilot is nonexistent, an estimate using a subsample selected through another sampling technique can be used instead. In the original paper describing the algorithm, the authors propose to use weighted case-control sampling with half the assigned sampling budget. For example, if the objective is to use a subsample with size <math> N=1000 </math>, first estimate a model <math>\tilde{\theta} </math> using <math> N_h = 500 </math> samples from weighted case control sampling, then collect another <math> N_h = 500 </math> samples using local case-control sampling.

=== Larger or smaller sample size ===

Line 34:

| volume=42

| issue=5

| pagepages=1693–1724

| ref=http://arxiv.org/abs/1306.3706}}</ref>

}}