Example Generation Under Constraints Using Cascade Correlation Neural Nets

AbstractHumans not only can effortlessly imagine a wide range of novel instances and scenarios when prompted (e.g., a new shirt), but more remarkably, they can adequately generate examples which satisfy a given set of constraints (e.g., a new, dotted, pink shirt). Recently, Nobandegani and Shultz (2017) proposed a framework which permits converting deterministic, discriminative neural nets into probabilistic generative models. In this work, we formally show that an extension of this framework allows for generating examples under a wide range of constraints. Furthermore, we show that this framework is consistent with developmental findings on children’s generative abilities, and can account for a developmental shift in infants’ probabilistic learning and reasoning. We discuss the importance of integrating Bayesian and connectionist approaches to computational developmental psychology, and how our work contributes to that research.


Return to previous page