Abstract

Transfer learning can be described as the distillation of abstract knowledge from one learning domain or task and the reuse of that knowledge in a related domain or task. In categorization settings, transfer learning is the modification by past experience of prior expectations about what types of categories are likely to exist in the world. While transfer learning is an important and active research topic in machine learning, there have been few studies of transfer learning in human categorization. We propose an explanation for transfer learning effects in human categorization, implementing a model from the statistical machine learning literature – the hierarchical Dirichlet process (HDP) – to make empirical evaluations of its ability to explain these effects. We present two laboratory experiments which measure the degree to which people engage in transfer learning in a controlled setting, and we compare our model to their performance. We find that the HDP provides a good explanation for transfer learning exhibited by human learners.