Step-by-step learning Articles uri icon

publication date

  • July 2025

start page

  • 1

end page

  • 12

volume

  • 284

International Standard Serial Number (ISSN)

  • 0957-4174

Electronic International Standard Serial Number (EISSN)

  • 1873-6793

abstract

  • The natural or generational learning process consists of building models based on available experiences. Each generation learns from the models obtained by its predecessors and obtains a new model for its own batch of experiences. In this paper, we discuss this step-by-step learning procedure for supervised classification and regression problems on large datasets. We show that the stepwise learning procedure performs competitively with respect to the approach that uses a single model for the entire dataset. This allows the step-by-step procedure to address larger datasets, and also, if necessary, respect the confidentiality of data from previous generations.

subjects

  • Computer Science
  • Statistics

keywords

  • batch processing; federated learning; machine learning; regression; supervised classification