The Cram Method for Efficient Simultaneous Learning and Evaluation

arXiv:2403.07031v1 Announce Type: new
Abstract: We introduce the "cram" method, a general and efficient approach to simultaneous learning and evaluation using a generic machine learning (ML) algorithm. In a single pass of batched data, the proposed method repeatedly trains an ML algorithm and tests its empirical performance. Because it utilizes the entire sample for both learning and evaluation, cramming is significantly more data-efficient than sample-splitting. The cram method also naturally accommodates online learning algorithms, making its implementation computationally efficient. To …

abstract algorithm arxiv cs.lg data evaluation general machine machine learning performance stat.co stat.me stat.ml tests trains type