Timezone: »

 
Towards Healing the Blindness of Score Matching
Mingtian Zhang · Oscar Key · Peter Hayes · David Barber · Brooks Paige · Francois-Xavier Briol

@
in NeurIPS 2022 Workshop on Score-Based Methods »
Event URL: https://openreview.net/forum?id=Ij8G_k0iuL »

Score-based divergences have been widely used in machine learning and statistics applications. Despite their empirical success, a blindness problem has been observed when using these for multi-modal distributions. In this work, we discuss the blindness problem and propose a new family of divergences that can mitigate the blindness problem. We illustrate our proposed divergence in the context of density estimation and report improved performance compared to traditional approaches.

Author Information

Mingtian Zhang (UCL)
Oscar Key (University College London)
Peter Hayes (University College London)
David Barber (University College London)
Brooks Paige (UCL)
Francois-Xavier Briol (University of Cambridge)

More from the Same Authors