Human-like Function Learning and Transfer

24.05.2017 - 17:00
24.05.2017 - 19:00

Pablo León-Villagrá

Doctoral Student, Institut für Language, Cognition and Computation; The University of Edinburgh, UK

"Humans can generalise in diverse ways that respect the abstract structure of a problem and can use knowledge in one context to inform decisions in another. Knowledge transfer is common in applied statistics, as when a practitioner recognises that kinds of regression problems involve certain parametric relationships. It is also at the heart of scientific progress, e.g. when analogies lead to new hypotheses and discoveries. In some situations, data are plentiful and transfer of knowledge is relatively unimportant; but when data are sparse, having appropriate prior knowledge is essential. Here. I will present empirical evidence that humans re-use abstract compositional properties of functions that they have learned previously, allowing them to extrapolate complex patterns from extremely sparse data. Furthermore, these learned compositional properties can be recombined allowing one-shot generalization to novel domains. Based on these results, I will argue for structured and compositional cognitive representations as a fundamental feature of human function learning."