Towards creativity characterization of generative models in the Activation Space

We’re going to be presenting some preliminary results in our work “Towards creativity characterization of generative models via group-based subset scanning” at Synthetic Data Generation Workshop at ICLR’21

Creativity is a process that provides novel and meaningful ideas. Current deep learning approaches open a new direction enabling the study of creativity from a knowledge acquisition perspective. Novelty generation using powerful deep generative models, such as Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs), have been attempted. However, such models discourage out-of-distribution generation to avoid instability and decrease spurious sample generation, limiting their creative generation potential. We propose group-based subset scanning to quantify, detect, and characterize creative processes by detecting a subset of anomalous node-activations in the hidden layers of generative models. Our experiments on original, typically decoded, and “creatively decoded” (Das et al., 2020) image datasets reveal that the proposed subset scores distribution is more useful for detecting creative processes in the activation space rather than the pixel space. Further, we found that creative samples generate larger subsets of anomalies than normal or non-creative samples across datasets. Also, the node activations highlighted during the creative decoding process are different from those responsible for normal sample generation.