Random Network Distillation as a Diversity Metric for Both Image and Text Generation

2020/10/13 by Liam Fowl, Fowl, Liam, Micah Goldblum +7
Biochemistry, Genetics and Molecular Biology · Computer Science · #Artificial Intelligence in Games #Cell Image Analysis Techniques #Computation and Language (cs.CL) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #cs.CL #cs.CV #cs.LG

paper · pdf · doi:10.48550/arxiv.2010.06715

arxiv created 2020/10/13 · openalex publication_date 2020/10/13 · arxiv updated 2020/10/15 · openalex created_date 2020/10/22 · openalex updated_date 2026/07/28

Abstract

Generative models are increasingly able to produce remarkably high quality images and text. The community has developed numerous evaluation metrics for comparing generative models. However, these metrics do not effectively quantify data diversity. We develop a new diversity metric that can readily be applied to data, both synthetic and natural, of any type. Our method employs random network distillation, a technique introduced in reinforcement learning. We validate and deploy this metric on both images and text. We further explore diversity in few-shot image generation, a setting which was previously difficult to evaluate.

Citations

• Adam: A Method for Stochastic Optimization
• ImageNet Large Scale Visual Recognition Challenge
• Pointer Sentinel Mixture Models
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Self-Attention Generative Adversarial Networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• GAN Augmentation: Augmenting Training Data using Generative Adversarial Networks
• Exploration by Random Network Distillation
• FIGR: Few-shot Image Generation with Reptile
• Jointly Measuring Diversity and Quality in Text Generation Models
• The generalization error of random features regression: Precise asymptotics and double descent curve
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Language Models are Few-Shot Learners

Related

• Measuring Diversity in Co-creative Image Generation
• Image Generation Diversity Issues and How to Tame Them
• Towards Mode Balancing of Generative Models via Diversity Weights
• GRADE: Quantifying Sample Diversity in Text-to-Image Models
• On the Diversity and Realism of Distilled Dataset: An Efficient Dataset Distillation Paradigm