Group-realizable multi-group learning by minimizing empirical risk
This addresses sample efficiency in multi-group learning for machine learning theory, but is incremental as it builds on known settings and methods.
The paper tackles the sample complexity of multi-group learning in the group-realizable setting, showing it improves over the agnostic setting even with infinite groups of finite VC dimension, but finds empirical risk minimization computationally intractable and suggests improper learning as an alternative.
The sample complexity of multi-group learning is shown to improve in the group-realizable setting over the agnostic setting, even when the family of groups is infinite so long as it has finite VC dimension. The improved sample complexity is obtained by empirical risk minimization over the class of group-realizable concepts, which itself could have infinite VC dimension. Implementing this approach is also shown to be computationally intractable, and an alternative approach is suggested based on improper learning.