Automated Machine Learning for Unsupervised Tabular Tasks
This addresses the challenge of automating machine learning for unlabeled data, though it is an incremental step toward broader model selection.
The paper tackles the problem of model selection for unsupervised tabular tasks like outlier detection and clustering by proposing LOTUS, which uses Optimal Transport distances to recommend pipelines based on dataset similarity, showing promising results against strong baselines.
In this work, we present LOTUS (Learning to Learn with Optimal Transport for Unsupervised Scenarios), a simple yet effective method to perform model selection for multiple unsupervised machine learning(ML) tasks such as outlier detection and clustering. Our intuition behind this work is that a machine learning pipeline will perform well in a new dataset if it previously worked well on datasets with a similar underlying data distribution. We use Optimal Transport distances to find this similarity between unlabeled tabular datasets and recommend machine learning pipelines with one unified single method on two downstream unsupervised tasks: outlier detection and clustering. We present the effectiveness of our approach with experiments against strong baselines and show that LOTUS is a very promising first step toward model selection for multiple unsupervised ML tasks.