Distribution Embedding Networks for Generalization from a Diverse Set of Classification Tasks
Abstract: We propose Distribution Embedding Networks (DEN) for classification with small data. In the same spirit of meta-learning, DEN learns from a diverse set of training tasks with the goal to generalize to unseen target tasks. Unlike existing approaches which require the inputs of training and target tasks to have the same dimension with possibly similar distributions, DEN allows training and target tasks to live in heterogeneous input spaces. This is especially useful for tabular-data tasks where labeled data from related tasks are scarce. DEN uses a three-block architecture: a covariate transformation block followed by a distribution embedding block and then a classification block. We provide theoretical insights to show that this architecture allows the embedding and classification blocks to be fixed after pre-training on a diverse set of tasks; only the covariate transformation block with relatively few parameters needs to be fine-tuned for each new task. To facilitate training, we also propose an approach to synthesize binary classification tasks, and demonstrate that DEN outperforms existing methods in a number of synthetic and real tasks in numerical studies.
Submission Length: Regular submission (no more than 12 pages of main content)
Changes Since Last Submission: Made changes to incorporate comments from the AE.
Assigned Action Editor: ~Tongliang_Liu1
License: Creative Commons Attribution 4.0 International (CC BY 4.0)
Submission Number: 260
Loading